Search

For businesses to operate effectively, networks and computer systems have become crucial instruments. They are now used in all professional fields, including the military, universities, banks, and insurance companies: as a result, a computer network enables a variety of tasks to be completed, including the sharing of data between users, the sharing of physical resources like the printer, the use of specific applications without having to install them on one's own computer, and the use of the Internet to conduct information searches, however security issues exist with computer networks. A computer would be more open to all types of attacks if it was connected to a network. Anomaly detection is the process of identifying outliers, or data points that differ significantly from the majority of other data points. In this paper, we present our solution in two chained phases, the first one is the Feature Selection Component using the combination of the two main algorithms ACO and Random Forest for the detection of relevant attributes, and the second phase is the Intrusion Detection Component based on the Isolation Forest algorithm to classify and detect intrusions. [ABSTRACT FROM AUTHOR]

Milk is one of the traditional foods in the human diet. However, subclinical mastitis infection in cows could compromise the nutritional composition of milk as well as the consumer safety. In this study, we investigated the possibility of implementing near infrared spectroscopy (NIRS), using a portable spectrometer, as a screening method to detect milk contaminated with subclinical mastitis. The spectra was acquired from a highly complex milk sample set (two seasons: winter and summer; two years: 2021 and 2022; from five geographical locations within Brazil: Goiás, Pará, Paraíba, São Paulo, and Santa Catarina; different feeding systems.) were analyzed using Principal Component Analysis (PCA) and Partial Least Squares Discriminant Analysis (PLS-DA), as well as machine learning models like Random Forest (RF) and Support Vector Machine (SVM). PCA revealed that the clustering of mastitis milk and non-mastitis milk samples was associated with variations in lactose content. This observation was also noted in PLS-DA, which achieved accuracy values of 78%, compared to the 62% achieved by RF and SVM in detecting mastitis milk. However, the sensitivity (recall) for detecting mastitis samples was higher for RF (78%), and SVM notably excelled in detecting non-mastitis milk (81%). Furthermore, it was demonstrated that the removal of outliers using Isolation Forest enhances the performance of RF and SVM models based on NIR spectra, with increases of up to 25% in the precision. In conclusion, the portable NIR spectrometer could potentially serve as a screening method in the dairy industry to detect mastitis milk samples. [Display omitted] • Milk with subclinical mastitis was classified by portable NIR spectrometer. • Lactose content is the main difference between mastitis and non-mastitis milk. • Isolation Forest improved 10–25% precision of RF and SVM by removing outliers. • RF reached a better sensitivity (recall) to classify mastitis milk samples. • PLS-DA reached the precision of 0.78 to classify milk with subclinical mastitis. [ABSTRACT FROM AUTHOR]

• We analyzed the isolation-forest-based methods' problem of lacking efficacy in selecting suitable hyperplanes to split data. • We proposed to use random projection techniques to improve isolation forest's efficacy in selecting suitable hyperplanes. • Experimental results on 30 real-world datasets show that the proposed methods outperform 12 SOTA outlier detectors. [Display omitted] Isolation Forest has a low computational complexity, hence has been widely applied to detect outliers in large-scale data. However, it suffers from the artifacts caused by the hyperplanes chosen, thereby failing to detect outliers in some specific regions. To tackle this problem, we propose the random-projection-based Isolation Forest, which works in two steps. First, we transform the data using the random projection technique. Then, we employ the Isolation Forest to identify outliers using the transformed data. Experimental results show that the proposed methods outperform 12 state-of-the-art outlier detectors. [ABSTRACT FROM AUTHOR]

Anomaly detection is nowadays a key data mining task. Anomaly detection methods generally look for patterns of "normal" profile and then identify data points that do not match that profile. One outstanding method, Isolation Forest, showed high detection performance and computational efficiency. This method does not profile normal instances. Instead, it seeks to separate "abnormal" data points that are "few and different", from the others. This method builds a forest of trees to define a model for classifying data points as normal or abnormal. Building trees involve recursive random partitioning of data. Instances resulting in shorter path lengths are considered anomalies because they are located in sparse regions whereas instances ending deeper in tree path are normal points. Most outlier detection approaches, including Isolation Forest, have been used on precise, crisp data. However, in real life, data may be imprecise, incomplete, ambiguous, or uncertain. In this paper, we propose a new approach, called Fuzzy Isolation Forest, which adapts the Isolation Forest method to fuzzy data to handle the "vagueness" of data. We use a fuzzy partitioning strategy based on the alpha-cut approach in which data instances will be assigned to tree nodes by comparing their membership value to an alpha-cut threshold. Fuzzy Isolation Forest is as efficient as Isolation Forest in detecting anomalies on fuzzy data without significantly increasing processing time. [ABSTRACT FROM AUTHOR]

Eurasian Curlew populations are declining in Europe despite conservation efforts. Mowing practices may attract Curlews to areas with a higher chance of survival, but this potential cannot be assessed due to limited documentation on mowing dates. This study developed a remote sensing method for mowing event detection by applying cloud masking, outlier detection via Isolation Forest, and data smoothing on satellite images to create a Normalised Difference Vegetation Index (NDVI) time series. GPS data from the LBV Society for the Protection of Birds and Nature in Bavaria was used to examine changes in Curlews' field use under mown and unmown conditions in their breeding areas. The developed approach detected 80 % of mowing events in trained data and 84 % in validation data with a ± three-day precision. Curlews visited fields significantly less often under unmown conditions and their field use increased substantially shortly after mowing events. Their reaction was stronger later in the season and is likely related to non-territorial behaviour. Fields under regulated mowing contracts showed more intensive Curlew activity than those conventionally managed. The workflow introduced for identifying mowing events through optical satellite imagery was designed with an emphasis on model robustness and on being accessible and reproducible for conservation practitioners and researchers. This simplified method successfully provided insights into factors influencing Curlews' use of grassland during their stay in their breeding areas. Mowing practices significantly impact their habitat choices, suggesting their use as an innovative conservation approach to recover Curlew populations. • Focus on mowing detection in agricultural grasslands and on field use of Eurasian Curlews. • An open-access mowing detection algorithm was developed focusing on practical use, based on remote sensing and ground truth data. • Curlews visited unmown fields significantly less and increased their field use shortly after mowing events. • Grasslands mown later in the season and fields under regulated mowing contracts showed more intensive Curlew activity. • We propose mowing regimes as active management tools to support the recovery of Curlew populations by guiding Curlews to more favourable regions in their breeding areas. [ABSTRACT FROM AUTHOR]

Photoplethysmography (PPG) signals, typically acquired from wearable devices, hold significant potential for continuous fitness-health monitoring. In particular, heart conditions that manifest in rare and subtle deviating heart patterns may be interesting. However, robust and reliable anomaly detection within these data remains a challenge due to the scarcity of labeled data and high inter-subject variability. This paper introduces a two-stage framework leveraging representation learning and personalization to improve anomaly detection performance in PPG data. The proposed framework first employs representation learning to transform the original PPG signals into a more discriminative and compact representation. We then apply three different unsupervised anomaly detection methods for movement detection and biometric identification. We validate our approach using two different datasets in both generalized and personalized scenarios. Our results demonstrate significant improvements: for movement detection, in the generalized scenario, AUCs improved from barely 0.5 to above 0.9 with representation learning. Importantly, inter-subject variability was substantially reduced, from around 0.4 to below 0.1. In the personalized scenario, AUCs became close to 1.0, with variability further reduced to below 0.05, indicating the effectiveness of both representation learning and personalization for anomaly detection in PPG data. Similar enhancements were observed in biometric identification, emphasizing how our approach can minimize inter-subject variability and enhance PPG-based health monitoring systems. • A two-stage unsupervised PPG anomaly detection using representation learning. • Showing the effect of person-specific models in enhancing the detection performance. • Validating the approach with two datasets in general and personalized scenarios. • Achieving performance improvements and reducing inter-subject variability. • Applying the framework to movement detection and biometric identification. [ABSTRACT FROM AUTHOR]

• We propose an innovative deterministic attribute selection for isolation forest. • We introduce algorithms based on imbalance, clustering, and non-linear transformations. • The proposed solutions have been tested on 25 real datasets. • Deterministic attribute selection for isolation forest enhances its effectiveness. Modern data mining techniques have been gained importance in recent years. In particular, anomaly detection algorithms, applied in key sectors of information technology, have been growing in popularity. One of the efficient and fast algorithms is Isolation Forest. The method consists of two separated stages: Forest formation and evaluation of elements. The first stage relies on forming a forest of isolation trees. Each tree is built in the same manner according to drawn samples and random divisions of data attributes. In this study, an innovative deterministic attribute selection method is proposed, maintaining its random value. New ideas based on imbalance, clustering, and a dispersion of values through non-linear transformation of elements are introduced and thoroughly analyzed. These novel anomaly detection approaches are applied to 25 real datasets, as well as our own artificially generated databases. The Area Under the ROC Curve and the Area Under the PR Curve are used as a measure of the outliers classification quality. The results of the numerical experiment have proven high efficiency and competitive evaluation speed of the proposals in comparison to other Isolation Forest-based approaches, as well as several other popular techniques. [Display omitted] [ABSTRACT FROM AUTHOR]

Electroencephalography (EEG) is commonly used for measuring brain activity information due to its high temporal resolution. However, it severely suffers from noises produced by non-brain sources, called EEG artifacts. Ocular, muscle, and power line artifacts are the most common ones. In this paper, a novel two-stage automatic artifact removal method is proposed to handle different artifacts under miscellaneous EEG applications. Empirical wavelet transform (EWT), canonical correlation analysis (CCA), and an outlier detection algorithm, isolation forest constitute the pipeline. In the first stage, EEG data are decomposed by CCA and preliminary purified by isolation forest. In the second stage, a further decomposition of the EEG data is conducted using EWT and CCA, followed by removal of irrelevant components using isolation forest. We thoroughly evaluate the qualitative and quantitative performance of the proposed method on multiple datasets. Experiment results show that the proposed method can effectively remove artifacts under complex conditions and different signal-to-noise ratios. Ablation studies and comparison results demonstrate the significance of the two-stage combination, which outperforms single-stage methods and state-of-the-art methods. This paper explores the potential for a fully data-driven and adaptive way for robust artifact removal in various EEG applications. • We combined empirical wavelet transform (EWT) with canonical correlation analysis and isolation forest. • We designed a two-stage removal process which focused on different artifacts at different stages. • We tested our method on multiple datasets, in a variety of ways. [ABSTRACT FROM AUTHOR]

• We propose a new unsupervised anomaly detection (AD) algorithm. • This algorithm is based on isolation forest with random hyperplanes instead of random dimensions. • The proposed method improves the existing extended isolation forest (EIF) in terms of computation time. This letter introduces a generalization of Isolation Forest (IF) based on the existing Extended IF (EIF). EIF has shown some interest compared to IF being for instance more robust to some artefacts. However, some information can be lost when computing the EIF trees since the sampled threshold might lead to empty branches. This letter introduces a generalized isolation forest algorithm called Generalized IF (GIF) to overcome these issues. GIF is faster than EIF with a similar performance, as shown in several simulation results associated with reference databases used for anomaly detection. [ABSTRACT FROM AUTHOR]

Anomaly detection is an important field in data science that has been widely researched and applied, generating many methods. Among these methods, the isolation forest algorithm is outstanding because of its efficiency and effectiveness, especially in regard to large-scale data. Unfortunately, this algorithm has some drawbacks, such as being unable to effectively handle local outliers, possibly leading to normal data masking the detection of outliers due to space partitioning along the coordinate axes, and low utilization of training data. To address these issues, in this paper we propose an improved isolation forest algorithm based on multilayer subspace dividing, named layered isolation forest, which adapts to the different distributions of the dataset by dividing the sample space into subspaces and evaluating the anomaly degree of data in different spatial ranges. This algorithm obtains a more accurate and reasonable anomaly score, avoids the problems of the original algorithm, and improves the performance metrics. According to the experimental results, the proposed method maintains the efficiency of the original algorithm and exhibits the best comprehensive performance compared with similar algorithms on artificial synthetic datasets and real-world datasets from multiple domains. • Isolation forest algorithm demonstrates limitations in isolating local anomalies. • Local anomalies are more easily isolated in subspace. • Combining global and local isolation forests enhances anomaly identification. • Layered Isolation Forest (LIF) overcomes traditional isolation forest limitations. [ABSTRACT FROM AUTHOR]

Understanding why some points in a data set are considered as anomalies cannot be done without taking into account the structure of the regular points. Whereas many machine learning methods are dedicated to the identification of anomalies on one side, or to the identification of the data inner-structure on the other side, a solution is introduced to answers these two tasks using a same data model, a variant of an isolation forest. The initial algorithm to construct an isolation forest is indeed revisited to preserve the data inner structure without affecting the efficiency of the outlier detection. Experiments conducted both on synthetic and real-world data sets show that, in addition to improving the detection of abnormal data points, the proposed variant of isolation forest allows for a reconstruction of the subspaces of high density. Therefore, the former can serve as a basis for a unified approach to detect global and local anomalies, which is a necessary condition to then provide users with informative descriptions of the data. [ABSTRACT FROM AUTHOR]

Lithology identification is a crucial task for reservoir characterization and evaluation. There exists an intricate non-linear response between formation lithology and logging data. However, it is difficult to avoid lithology mislabeling due to human error and interpretation coarsening, and label quality can seriously affect the effectiveness of supervised learning. The presence of noisy labels makes it essential to learn with noisy labels. Noise-filtering methods and noise-robust algorithms only concentrate on a singular aspect of data or algorithm. In this paper, hybrid noise label filtering and correction framework for lithology identification (HNFCL) is proposed. Isolation forest is utilized to detect suspicious data, as it is efficient and fast. Baseline classifiers are built by ensemble tree models. In particular, the labels of abnormal data are removed and Tri-training semi-supervised method is introduced to relabel these data, which minimizes the loss of valid training data. Comprehensive experiments of the HNFCL framework, noise filtering methods and deep neural network methods with optimized loss functions were carried out in the industrial application of logging lithology identification. HNFCL achieved average accuracy of 87.94% and 94.93% in two study wells. These results outperformed the noise filtering methods and showed no significant difference from the state-of-the-art method. The correction of noise by HNFCL will provide a prospect for lithology identification applications. [ABSTRACT FROM AUTHOR]

• Non-vibration signals directly derived from drive-end (current and speed) of ball screw are adopted and fused together to constitute multi-source signal samples of DTL models. • An unsupervised iForest algorithm is delicately designed to generate pseudo-labels of samples automatically by detecting anomalies on benchmark-comparison datasets. • Fine-tuning strategy is introduced into the proposed ball screw transfer fault diagnosis, several experimental results verify the feasibility and superiority of the proposed method. Vibration-based fault diagnosis methods of ball screw are susceptible to noise and transmission path. Moreover, the accuracy of supervised deep learning models depends on large amounts of labeled samples, which are not only difficult to obtain but also laborious to label. Therefore, to solve these problems, a zero-cost unsupervised transfer method based on non-vibration signals fusion is proposed to achieve ball screw fault diagnosis in this paper. Firstly, non-vibration signals, such as current and speed, are adopted and orderly fused together to constitute multi-source fusion signal samples, which are easier to obtain and contain fewer interferences than vibration signals. Secondly, by virtue of its excellent abnormal detection ability, isolation forest algorithm is circularly utilized to generate pseudo-labels of source domain samples without manual labeling, which further realizes zero-cost sample labeling and unsupervised process. Finally, large amount of generated pseudo-labeled samples of source domain is applied to pre-train the transfer model parameters, and fine-tuning strategy with small number of labeled samples of target domain is used to complete transfer fault diagnosis of ball screw. The effectiveness of the proposed method is verified by ball screw signals across three different operation conditions, ablation and comparison analysis are also studied to illustrate its advantages. [ABSTRACT FROM AUTHOR]

Although deep neural networks (DNNs) have performed well on many perceptual tasks, they are vulnerable to adversarial examples that are generated by adding slight but maliciously crafted perturbations to benign images. Adversarial detection is an important technique for identifying adversarial examples before they are entered into target DNNs. Previous studies that were performed to detect adversarial examples either targeted specific attacks or required expensive computation. Designing a lightweight unsupervised detector is still a challenging problem. In this paper, we propose an A uto E ncoder-based A dversarial E xamples (AEAE) detector that can guard DNN models by detecting adversarial examples with low computation in an unsupervised manner. The AEAE includes only a shallow autoencoder that performs two roles. First, a well-trained autoencoder has learned the manifold of benign examples. This autoencoder can produce a large reconstruction error for adversarial images with large perturbations, so we can detect significantly perturbed adversarial examples based on the reconstruction error. Second, the autoencoder can filter out small noises and change the DNN's prediction on adversarial examples with small perturbations. It helps to detect slightly perturbed adversarial examples based on the prediction distance. To cover these two cases, we utilize the reconstruction error and prediction distance from benign images to construct a two-tuple feature set and train an adversarial detector using the isolation forest algorithm. We show empirically that AEAE is an unsupervised and inexpensive detector against most state-of-the-art attacks. Through the detection in these two cases, there is nowhere to hide adversarial examples. • We observe that adversarial detection is sensitive to the perturbation level. • We train a shallow autoencoder to find two key features from adversarial examples. • We propose a lightweight and unsupervised adversarial detector. [ABSTRACT FROM AUTHOR]

Bad channels result from the inherent instability of wearable optically pumped magnetometers (OPMs). These unreliable channels significantly impede subsequent signal analysis. The signal space separation method for OPMs, which relies on time-domain features such as the z-score, cannot accurately identify these unreliable channels. We introduce a novel approach for automatic bad channel identification in OPMs, utilizing frequency domain features. Box-plot technique enables predicting the number of unreliable channels. The predicted ratio of unreliable channels is incorporated as a crucial parameter in the Isolation Forest algorithm to achieve the automated identification of bad channels. To evaluate the effectiveness of our method, we employed three accuracy metrics and conducted comparative assessments via the following methods: box-plot, density-based spatial clustering of applications with noise, and local outlier factor. Our approach is validated via simulation experiments and real measurements using OPM-magnetoencephalography data. The efficiency of subsequent OPM-MEG preprocessing and source location analysis is improved. • At present, bad channel identification mainly relies on manual experience. • The SSS for OPM cannot achieve satisfactory accuracy in identifying bad channels. • The proposed method is verified to be effective. [ABSTRACT FROM AUTHOR]

Music, a significant and exquisite part of human culture, owns abundant features and enjoys a long-standing history. Music evolves in society over time, while artists' music gets influenced by personal experiences, external events, and inspirations from predecessors. In this paper, we propose a model named MFCSNet that measures musical influence by utilizing the data sets of musical characteristics and links between music influencers and followers. MFCSNet applies multiple indicators and has more analysis perspectives, and well reflects the influence of different kinds of music in various fields. Firstly, we analyze the influencer–follower relations by looking at the network of musical influence, observing the correlation between followers and influencers, and closely examining several sub-networks extracted from the entire network. Secondly, we propose measures that quantify the similarities within and between musical genres, using musical characteristics, such as danceability, energy, and valence, in order to measure the influence between artists and find the more influential characteristics. Furthermore, we apply MFCSNet on the whole timeline to analyze the evolutions and revolutions of music through time, with the goal of revealing the relation between music and culture, society, politics, and technologies. • Analyzed music evolution & influence through diverse methods. • Built an influencer-follower social network, revealing a small-world artist effect. • Identified a scale-free network with output degree nodes showing influence. • Confirmed findings in sub-networks: New Age and Jazz music. [ABSTRACT FROM AUTHOR]

Machine learning technology has been widely applied in the field of seismic interpretation. In most cases, machine learning assisted seismic interpretation is calibrated or constrained by wells. However, due to the limitation of drilling cost, sometimes there are only a few samples can be obtained from the well-points for a specific layer, which is insufficient to guarantee the generalization ability of supervised learning. In this article, we propose a novelty semi-supervised method by combining the unsupervised isolation forest with split-selection criterion (SCiForest) algorithm and the supervised feature selection process together. The key of the proposed method is to be able to make full use of both the self-contained distribution information of multiple seismic attributes and the calibration information of limited well-points at the same time. To highlight the advantages of the proposed method referred to conventional supervised and unsupervised methods, we take the channel identification practice in the western Bohai Sea as a case study for comparison. Further discussion confirms that the proposed method can improve the visibility of channel effectively by fusing the relevant information in amplitude, frequency, and morphological attributes with limited calibration, which may provide a reliable alternative way for further machine learning assisted seismic interpretation. • A novelty semi-supervised method which combines the unsupervised SCiForest and supervised attribute selection is proposed. • Comparison confirms the advantage of the proposed method in seismic channel identification referred to existing methods. • We reveal that the limited well-point samples are insufficient for supervised learning based seismic interpretation. • Utilizing distribution patterns in seismic attributes can supplement the pattern recognition for channel identification. [ABSTRACT FROM AUTHOR]

While there have been extensive studies in the health assessment of the bearings using the vibration signal, most have focused on the constant operating conditions. The field, however, operates often under time-varying conditions as is the case of the automotive wheel bearing. In this study, a novel health indicator (HI) is proposed to address this problem based on the Isolation Forest algorithm, which was originally developed for anomaly detection. The method is advantageous in two aspects: the HI is not influenced by the type of operating conditions whether it is constant or time-varying. Only the data under normal condition are used to construct the HI without the need of run-to-failure data. The method is demonstrated by the three cases with different types of bearing and operating conditions ranging from constant to the highly variable conditions. As a result, monotonic trends are obtained for the HI in all cases, which may be useful for the prognostic monitoring. Furthermore, in comparison with the HI constructed by the run-to-failure data in the previous literature, it is found that the trends of HI agree reasonably well with each other, which supports the validity of the method. [ABSTRACT FROM AUTHOR]

Modern data analysis allows to infer from the data to a very large extent. Particularly noteworthy are techniques based on the detection of anomalies in datasets. Unsupervised approaches with no labeled records are gaining ground. One of the most frequently used and efficient unsupervised methods is Isolation Forest (IF) with its numerous specialized modifications. The considerations of this paper are devoted to the enhancement of Minimal Spanning Tree-Based Isolation Forest (MSTBIF). MSTBIF exhibits high effectiveness. Its evaluation function consists of the sum of two parameters. Hence, the main motivation of the work is a better composition of the MSTBIF evaluation function parameters. Crucial improvements are made in two directions. Firstly, the output parameters related to the tree height and the distance to the nearest tree leaf are normalized. Normalizations take place in a new manner. The properties of created minimal spanning trees are used. Secondly, a simple aggregation of output parameters has been modified. The work introduces an innovative anomaly evaluation scheme. Namely, fuzzy rules prepared in the Takagi-Sugeno inference model are applied. Novel proposal allows for a more precise inference about the occurrence of anomalies in the considered elements. In order to compare the results, the AUC values are measured. A series of experiments with IF, MSTBIF, and new approach are carried out. The research results on the new proposal prove a significant refinement of the anomaly assessment. The improvement over MSTBIF is clearly noticeable. Compared to the basic IF method, the results are better by about 6%. [Display omitted] • We propose innovative score evaluation function for MSTBIF method. • We introduce fuzzy rules in the Takagi-Sugeno model for score evaluation. • The proposed solution has been tested on 26 real world datasets. • The results of MSTBIF-TS indicate its high effectiveness in anomaly detection. [ABSTRACT FROM AUTHOR]

One of the main ingredients to learn a visual representation of an object using the Convolutional Neural Networks is a large and carefully annotated dataset. Acquiring a dataset in a demanded scale is not a straightforward task; therefore, the community attempts to solve this problem by creating noisy datasets gathered from web sources. In this paper, this issue is tackled by designing a vehicle recognition system using Convolutional Neural Networks and noisy web data. In the proposed system, the transfer learning technique is employed, and behavior of several deep architectures trained on a noisy dataset are studied. In addition, the external noise of the gathered dataset is reduced by exploiting an unsupervised method called Isolation Forest, and the new training results are examined. Based on the experiments, high recognition accuracies were achieved by training two states of the art networks on the noisy dataset, and the obtained results were slightly improved by using the proposed noise reduction framework. Finally, a demonstration application is provided to show the capability and the performance of the proposed approach. [ABSTRACT FROM AUTHOR]

We survey unsupervised machine learning algorithms in the context of outlier detection. This task challenges state-of-the-art methods from a variety of research fields to applications including fraud detection, intrusion detection, medical diagnoses and data cleaning. The selected methods are benchmarked on publicly available datasets and novel industrial datasets. Each method is then submitted to extensive scalability, memory consumption and robustness tests in order to build a full overview of the algorithms’ characteristics. [ABSTRACT FROM AUTHOR]

The development of efficient and interpretable anomaly detection systems is fundamental to keeping production costs low, and is an active area of research in semiconductor manufacturing, particularly in the context of using Optical Emission Spectroscopy (OES) data. The high dimension and correlated nature of OES data can limit the performance achievable with anomaly detection systems. In this paper we present a dimensionality reducing variable selection and isolation forest based anomaly detection and diagnosis methodology that addresses these issues. In particular, it takes account of isolated variables that can be overlooked when using conventional approaches such as PCA, and provides greater interpretability than afforded by PCA. The proposed methodology is illustrated with the aid of simulated and industrial plasma etch case studies. [ABSTRACT FROM AUTHOR]

In an Industry 4.0 ecosystem, all the essential components are digitally interconnected, and automation is integrated for higher productivity. However, it invites the risk of increasing cyber-attacks amid the current cyber explosion. The identification and monitoring of these malicious cyber-attacks and intrusions need efficient threat intelligence techniques or intrusion detection systems (IDSs). Reducing the false positive rate in detecting cyber threats is an important step for a safer and reliable environment in any industrial ecosystem. Available approaches for intrusion detection often suffer from high computational costs due to large number of feature instances. Therefore, this paper proposes a novel unsupervised IDS for Industry 4.0 which we term as: Unsupervised Intrusion Detection System for Industry 4.0 (UInDeSI4.0). We have substantiated the proposed UInDeSI4.0 approach through its experimentation on the well-known UNSW-NB15 Industry 4.0 dataset. The proposed UInDeSI4.0 employs feature selection approaches to obtain minimal and optimal features. These features are then used to train isolation forest to detect network traffic threats in an unsupervised manner. Accordingly, the proposed UInDeSI4.0 approach can efficiently differentiate between the normal events and the attacks or intrusions in environments with no label information. Experimental results show that the proposed UInDeSI4.0 provides better accuracy (∼ 63%) and a minimal feature set (nine) compared to traditional IDSs. In contrast to deep learning approaches, UInDeSI4.0 generates faster results with minimum features. In conclusion, we establish the superiority of UInDeSI4.0 approach as an accurate and computationally efficient IDS for Industry 4.0. [ABSTRACT FROM AUTHOR]

[Display omitted] • Isolation forests (IFs) are used to detect outliers in coal characterisation and processing data. • IFs detect outliers that were not visible with principal component score plots. • A random forest (RF) was used to relate plant data with IF anomaly scores. • A kernel SHAP approach was used to plausibly explain the predictions of the random forest. • Variable importance measures based on the RF impurity criterion were most similar to tree SHAP for global feature analysis. Modelling the characteristics and composition of coal is important, as proximity data and other measurements to do so are typically expensive or hard to acquire in real-time. Understanding anomalies in these relatively small data sets are important, as removal may result in an unnecessary loss of data or bias in the data used in the model. Although anomaly detection has been considered in-depth in the literature, very little work has been devoted to the explanation of anomalies. In this paper, a general anomaly detection and identification methodology is considered, based on three models, viz an isolation forest, a random forest and a tree SHAP explanatory model. Three case studies related to the composition of coal and coal processing are considered. In these case studies, the IF-RF-SHAP approach identified outliers of data anomalies not identifiable with principal component analysis. The model is a new variant of some of the integrated approaches that have recently been considered. Further contribution of the study lies in the empirical comparison of IF anomaly scores with distance-based and reconstruction-based anomaly scores generated with principal component models. In the case studies considered, the IF anomaly scores were better able to identify anomalies in the data than the scores derived from the principal component models. As a result, the methodology can complement distance-based approaches, such as principal component analysis, to explain anomalies or outliers detected in data. Apart from the proposed IF-RF-SHAP approach, four approaches to compare the contributions of variables in random forest models are considered as well. These were simple correlation of individual predictors with anomaly scores of samples, random forest prediction based on an impurity criterion, random forest prediction based on a permutation criterion, as well as the tree SHAP approach. If the latter is considered as a benchmark, then the impurity criterion gave the most reliable results, while simple predictor correlations gave the least reliable results. [ABSTRACT FROM AUTHOR]

Households and businesses are becoming increasingly reliant on IoT systems since they can reliably automate many tasks that otherwise require extensive human involvement. As a result, the number of IoT devices that are being added to the Internet is increasing steeply. To cater to the increasing demand and attract customers, manufacturers are focusing on producing cheaper and easy-to-use IoT devices without giving enough importance to the security aspect. With the recent developments in machine learning algorithms, IoT device manufacturers have started using supervised intrusion detection models to protect IoT systems from security breaches. However, these systems are being targeted by novel attacks that cannot be detected by the already deployed supervised models. Thus, it is important to develop solutions to protect the IoT networks/systems from these previously unknown intrusions. MQTT is a widely used network protocol in IoT systems due to its lightweight and flexible nature. In this paper, we propose a novel unsupervised GAN and autoencoder-based model, called GAN-AE, for detecting unknown intrusions in MQTT IoT applications. The performance of the proposed GAN-AE model is compared with other popular unsupervised models namely autoencoder, One-Class SVM (OCSVM), and Isolation Forest (IF). The GAN-AE model performed superior to other models with accuracy, F1-Score of 0.97 on our custom-built and public MQTT dataset. [ABSTRACT FROM AUTHOR]

Anomaly Detection is an unsupervised learning task aimed at detecting anomalous behaviors with respect to historical data. In particular, multivariate Anomaly Detection has an important role in many applications thanks to the capability of summarizing the status of a complex system or observed phenomenon with a single indicator (typically called 'anomaly score') and thanks to the unsupervised nature of the task that does not require human tagging. The Isolation Forest is one of the most commonly adopted algorithms in the field of Anomaly Detection due to its proven effectiveness and low computational complexity. A major problem affecting Isolation Forest is represented by the lack of interpretability, an effect of the inherent randomness governing the splits performed by the Isolation Trees, the building blocks of the Isolation Forest. In this paper, we propose effective yet computationally inexpensive methods to define feature importance scores at both global and local levels for the Isolation Forest. Moreover, we define a procedure to perform unsupervised feature selection for Anomaly Detection problems based on our interpretability method. Such a procedure also serves the purpose of tackling the challenging task of feature importance evaluation in unsupervised anomaly detection. We assess the performance on several synthetic and real-world datasets, including comparisons against state-of-the-art interpretability techniques, and make the code publicly available to enhance reproducibility and foster research in the field. [ABSTRACT FROM AUTHOR]

• An approach for detection of damper malfunction is developed by implementing isolation forest algorithm on characterized cumulative displacement measurements. • Physics-informed correlation between the temperature variation and the displacement measurements are utilized for identifying measurement noise to ensure reliable datasets for malfunction detection. • Characterization with hourly cumulative displacement can significantly reduce the data complexity while capturing the intrinsic data characteristics that influence the bridge behavior. • The approach was verified with real monitoring data in a suspension bridge by detection of malfunction in the viscous damper. Long-span cable-supported bridges are flexible and prone to significant deformations under temperature, wind, and vehicle loads. Such large deformations induce frequent longitudinal movement at the end of the bridge girder, which can cause malfunction or failure in the components such as expansion joints, bearings, or longitudinal viscous dampers. This paper develops a condition assessment approach using the girder end displacement measurements, which primarily includes four steps, namely, data preprocessing, feature characterization, datasets preparation and splitting, and anomaly detection. The data preprocessing step considers the physics-based correlation between the temperature and displacement, which is used to pre-check noisy data issues in the displacement monitoring data. Subsequently, hourly cumulative displacement is calculated as the feature to feed into the machine learning model. The third step prepares the datasets and splits them into training and testing datasets. Then the isolation forest algorithm is implemented for anomaly detection with well-tuned parameters of the number of decision trees and the number of samples in each decision tree. Structural health monitoring data of a suspension bridge are collected to verify the proposed approach, and the detection targets are specified as viscous damper malfunction and a specific holiday period. The proposed approach demonstrates its ability to detect both events successfully, which can shed light on the predictive and preventive maintenance of the bridge. [ABSTRACT FROM AUTHOR]

Anomaly detection is often used to identify and remove outliers in datasets. However, detecting and analyzing the pattern of outliers can contribute to future business decisions or increase the accuracy of a learning algorithm. Selecting the applicable outlier detection method for a dataset requires human intervention and analysis due to the challenge of choosing an efficient technique suitable for all types of attributes. This work presents a procedure for anomaly detection and analysis. The procedure is feature-wise (i.e., processes each feature independently), uses T different anomaly detection techniques (for T > 1), and estimates the best technique using predefined thresholds. It is a generic method that does not depend on the model type and can be applied to supervised and unsupervised learning. In addition, this method does not impute or remove the outliers, as they should be adapted according to the dataset or business requirements. The significant advantage of this method is the ability to use different techniques to detect anomalies since it is applied per feature and not per record, as in traditional anomaly detection methods. Furthermore, the method uses a new measure, Noise Ratio (NR), which describes the level of agreement between our method's result and traditional anomaly detection techniques. The results showed that all the compared techniques identified non-anomalous features with consistent results between the various algorithms. In the proposed method, NR found up to 20% of the non-anomalous values marked as outliers and improved up to 10% in finding outliers in datasets compared to traditional anomaly detection algorithms. [ABSTRACT FROM AUTHOR]

• Correlation analysis of water level data in explorational borehole groups. • Multivariate LSTM model for predicting the variation of water level. • IForest model for detecting water level variation anomalies. • Model calibration based on empirical threshold setting method. • Water inrush risk evaluation and early warning based on the coupled model of LSTM + IForest. Water level variation of explorational boreholes in mining sites is one of the most direct representations of water inrush risk. Despite recent efforts on mine water inrush accident prediction based on various types of observation data including water level of boreholes using a wide range of machine learning models, the accuracy and timeliness of the prediction for major accidents are still unsatisfactory. In this study, a novel prediction method based on the analysis of water level variation anomalies in borehole groups is proposed using the coupled model of long short-time memory (LSTM) and Isolation forest (IForest). Multi-variate LSTM algorithm is firstly used to model the correlated time sequence water level data in the borehole group, followed by the full extraction of the data variation feature based on the combination of variables including the prediction error being used as inputs of the IForest algorithm. As a result, anomalous variations of the data are detected as possible precursors of water inrush accident. Finally, a set of water inrush risk level grading rule is established based on the number of simultaneous anomalous data instances being detected within a borehole group. The results obtained for the water inrush accident in Xingdong coal mine demonstrate extraordinary timeliness and high accuracy as high water-inrush risk warning is issued for more than 2 days earlier than the time of actual sighting of the accident with very few false alarms being caused throughout the period of the testing dataset. This study provides an innovative solution to the low prediction accuracy and slow responding speed during major water inrush accidents of existing methods. The major parameters of the proposed coupled model of LSTM + IForest can also be easily calibrated based on traditional empirical methods for the best performance, allowing the method to be widely applied in various mining conditions. [ABSTRACT FROM AUTHOR]

• Established a database with 1375 FRP-concrete direct shear test specimens. • Analyzing the correlation of six variables revealed that five factors affecting the FRP-concrete interfacial shear capacity most. • Six machine learning models substantially increased the prediction accuracy compared with the sixteen traditional empirical equations, and they especially reduced the variation. • Based on the ANN algorithm, an accurate, explicit and practical equation was derived to predict the FRP-concrete interfacial shear capacity. Externally bonding fiber reinforced polymer (FRP) to concrete structures is an effective way to enhance the mechanical performance of concrete structures. Many equations have been proposed to predict the interfacial bond strength for FRP-concrete structures but have limited accuracy due to the complexity of the bond behavior. This study proposes to formulate the FRP-concrete interfacial bond strength based on machine learning (ML) methods, which have emerged as a promising alternative to achieve high prediction accuracy in high-dimension problems. To this end, a database containing 1,375 FRP-concrete direct shear test specimens that failed due to interfacial debonding was established. The database was improved using an unsupervised isolation forest that identified and eliminated anomalous data, and was then used to train six ML models, namely artificial neural networks (ANN), support vector machine, decision tree, gradient boosting decision tree, random forest, and XGboost algorithms, to predict the FRP-concrete interfacial bond strength. The ML predictive models showed higher accuracy than 16 existing equations in the literature. The XGBoost model showed the highest accuracy, and its coefficient of variation was 54% lower than the existing equation with the highest accuracy among those considered. The ANN model was used to perform a parametric study on the influencing parameters, and a new equation was generated to predict the interfacial bond strength, considering the key influencing parameters. The equation enables interpretation of the ML models. The study combines ML models and traditional physical models to achieve a novel, interpretable ML method for predicting FRP-concrete interfacial bond strength. [ABSTRACT FROM AUTHOR]

• Propose a wheelchair fall detection system based on Isolation Forest Algorithm. • Propose a ubiquitous architecture for the fall detection system. • Perform Feature engineering and the selection of most appropriate features to train the classifier achieving high accuracy. • Utilize the data from low-cost embedded sensors to capture data patterns for fall cases. Falls represent one of the major health risk issues world-wide. In this paper, a wheelchair fall anomaly detection framework based on the hybrid Isolation Forest (IF) and threshold based method (TBM) is proposed. The sensor data was obtained from tri-axial orthogonal accelerometer and gyroscope MPU-6050 sensor. In order to handle the uncertainties due to the sensor noise factor, a method based on Zero Angular Rate Update (ZART) and Complementary Filter (CF) was utilized. The concept of sensor fusion was applied to create multi-dimensional training data from the best features selected using the ReliefF algorithm. The IF detector was trained on the best eight-dimensional features selected from 48 features for the detection of the fall events in contrast to 38 features defined in our previous publication in Sheikh and Jilani. Further, the design of a ubiquitous architecture incorporating the proposed fall detection scheme was proposed. The g-mean score accuracy was achieved up-to ∼97.1% with the proposed IF +threshold hybrid strategy. While using only the IF method, an F1-score of 96.2%, an AUC-ROC score of 95.3% and a g-mean score of 92.9 % was obtained. [ABSTRACT FROM AUTHOR]

• Predictive modeling of stock price momentum of six Indian Companies is proposed. • The framework is based on Navier-Stokes equation. • The forecasting framework integrates iForest, BorutaShap, and Facebook's Prophet. • The neoteric framework outperformed benchmarking models. • Explainable AI framework is used to comprehend the influence of features. The paper presents a framework for forecasting stock prices for normal time and pandemic period and tests its forecasting efficacy. It is based on the specification of dependent and independent variables drawn from the Navier-Stokes equation, which defines the relationship between velocity and kinematic viscosity, density, pressure, and external force field. We define stock market proxies for the variables used in the equation in modeling stock price behavior and extend our framework to build in aspects of behavioral finance like hindsight bias, representative bias, and anchoring bias. The predictive exercise has been carried out on stock prices of Indian companies belonging to different industrial verticals to test whether our framework can handle diversity. AI framework comprising Isolation Forest (iForest), BorutaShap, and Facebook's Prophet Algorithm has been used for automatic anomaly elimination, dedicated feature screening, and forecasting future stock price movements. The results reveal interesting insights into the varied response of stock prices of different companies to macroeconomic shocks and would be quite useful for fund managers, regulators, and policymakers. Several numerical and statistical checks indicate the efficiency of our proposed framework in yielding superior quality forecasts, both in normal and challenging times. [ABSTRACT FROM AUTHOR]

Vehicle trajectory data derived from automatic vehicle location (AVL) and automatic vehicle identification (AVI) systems provide critical support for intelligent transportation systems. However, the field-obtained vehicle trajectories are usually incomplete due to sensor malfunction or communication issues. To recover the incomplete data, the existing reconstruction methods have to impose strong assumptions on driver route choice behaviors (network level) and/or traffic dynamics (link level). With the tremendous data available, leveraging data-driven approaches to address the vehicle trajectory reconstruction problem with minimal assumptions is promising. This paper proposes a general dynamic sequential learning framework to reconstruct vehicle trajectory points for both AVL and AVI data. First, an Isolation Forest based ensemble learning model is developed to extract trajectory sequences attributed to different trips in an entire trip chain of a vehicle. Second, the dynamic recurrent neural network (dynamic RNN) is tailored to learn the underlying patterns from the complete AVL and AVI trajectories, respectively. Third, a sequential prediction scheme is customized to reconstruct AVL and AVI trajectories based on the trained networks. To validate the proposed method, two experiments are conducted. One is a simulation experiment with AVL data gathered from a well-calibrated simulation model. The other is a field experiment with AVI data collected from a real-world automatic license plate recognition (ALPR) system. The results show that the proposed method achieves superior performance for both AVL and AVI data compared with the traditional methods. The impacts of different sampling rates and traffic conditions on the model performance are also discussed. • A general learning framework for vehicle trajectory reconstruction is developed. • An Isolation Forest based model is adopted to extract trip trajectory sequences. • The dynamic RNN is tailored to learn the patterns from the complete trajectories. • A sequential prediction scheme is customized to reconstruct incomplete trajectories. • Two comparisons are conducted to demonstrate the effectiveness of the framework. [ABSTRACT FROM AUTHOR]

A novel hybrid Garra rufa Fish optimization (GRFO) – isolation Forest (iForest) soft computing approach is proposed in this paper for optimizing the controller parameters in an isolated test microgrid. The test microgrid comprises of two PV units and one wind generator synchronized via Voltage Source Converters (VSC) in the ac side. Each VSC is regulated based on power generation in individual micro sources. The operation of VSC is the key factor maintaining the stability of the micro grid. Each VSC is controlled by outer power loop and inner current regulation loop employing PI controllers. The performance of the VSC is enhanced using the proposed hybrid GRFO – iForest algorithm for tuning the PI controller gains. The performance of the proposed approach is then compared with conventional tuning PI controller and optimization techniques like Ant Lion Optimization (ALO) and Modified Ant Lion optimization based Artificial Neural Network (MALANN). The gain parameter of the proposed controller is optimally tuned and the controller provides reliable sustainable microgrid system operation. The proposed approach is simulated in MATLAB/Simulink and validated using OP-4500 Real Time Simulator environment. Based on the results, the performance of the GRFO-iForest approach is superior during steady-state and transient operation. It enhances the sustainability of the microgrid by restoring normal operating conditions after a small physical disturbance and the microgrid remains stable with the optimized regulator. The stability of the overall system is proved with the mathematical model. • A hybrid optimization technique is proposed to optimize the gains of PI controller. • The control of Voltage Source Converter (VSC) is optimized in accordance with the available generation from renewable energy sources. • The optimized controller stabilizes the output power of the microgrid generators effectively under various operating conditions. • The steady-state error on the power output of distributed generators is reduced by 88% compared to other methods. • A sustainable microgrid system is developed which is fault tolerant in various operating conditions. [ABSTRACT FROM AUTHOR]

To conduct preemptive essential maintenance, predictive maintenance detects the risk of unexpected shutdowns in a manufacturing system, thereby ensuring operational continuity. Traditional methods that heavily rely on the domain knowledge of expert engineers to detect any abnormal status in processing facilities are extremely time-consuming and domain-dependent. Conversely, recently studied data-driven approaches without much domain knowledge have yielded fairly good performance. However, most only identify whether the current status is normal or abnormal and do not offer any explanations or analyses. In this paper, we propose a real-time explainable anomaly detection framework for predictive maintenance in a manufacturing system. Various well-known anomaly detection algorithms are investigated to construct a framework suitable for shutdown prognosis. In addition, model interpretation techniques are also employed to provide a reasonable explanation for a detected shutdown. The experimental results on a real-world dataset derived from a chemical process show that the proposed framework could identify abnormal signs early and derive significant causes for each detected shutdown. • We proposed explainable anomaly detection framework for predictive maintenance. • The framework detects the warning signs of a shutdown early in real-time. • The framework detects anomalies using only the characteristics of normal instances. • The framework provides the root causes of a detected shutdown. [ABSTRACT FROM AUTHOR]

• We show that GMM are efficient for the joint reconstruction of vegetation indices. • A robust GMM is proposed to take into account the presence of irrelevant samples. • The proposed approaches are validated on wheat and rapeseed parcels in France. • Using Sentinel-1 data can improve the imputation of vegetation indices. • The proposed algorithms are applied to the detection of anomalous crop development. Missing data is a recurrent problem in remote sensing, mainly due to cloud coverage for multispectral images and acquisition problems. This can be a critical issue for crop monitoring, especially for applications relying on machine learning techniques, which generally assume that the feature matrix does not have missing values. This paper proposes a Gaussian Mixture Model (GMM) for the reconstruction of parcel-level features extracted from multispectral images. A robust version of the GMM is also investigated, since datasets can be contaminated by inaccurate samples or features (e.g., wrong crop type reported, inaccurate boundaries, undetected clouds, etc). Additional features extracted from Synthetic Aperture Radar (SAR) images using Sentinel-1 data are also used to provide complementary information and improve the imputations. The robust GMM investigated in this work assigns reduced weights to the outliers during the estimation of the GMM parameters, which improves the final reconstruction. These weights are computed at each step of an Expectation–Maximization (EM) algorithm by using outlier scores provided by the isolation forest (IF) algorithm. Experimental validation is conducted on rapeseed and wheat parcels located in the Beauce region (France). Overall, we show that the GMM imputation method outperforms other reconstruction strategies. A mean absolute error (MAE) of 0.013 (resp. 0.019) is obtained for the imputation of the median Normalized Difference Index (NDVI) of the rapeseed (resp. wheat) parcels. Other indicators (e.g., Normalized Difference Water Index) and statistics (for instance the interquartile range, which captures heterogeneity among the parcel indicator) are reconstructed at the same time with good accuracy. In a dataset contaminated by irrelevant samples, using the robust GMM is recommended since the standard GMM imputation can lead to inaccurate imputed values. An application to the monitoring of anomalous crop development in the presence of missing data is finally considered. In this application, using the proposed method leads to the best detection results, especially when SAR data are used jointly with multispectral images. Exploiting the information contained in cloudy multispectral images instead of removing these images is beneficial for this application. [ABSTRACT FROM AUTHOR]

During the operation of nuclear power system, the instrument detection data often deviates from the normal operating state for a short time due to system or environmental fluctuations. And the control system will send an alarm signal, resulting in the false alarm. Aiming at the problem of false alarm, three algorithms are improved and combined to form a false alarm algorithm in this paper. The algorithm consists of the transient operating parameters processing algorithm and the nuclear power system anomaly identification algorithm. The transient operating parameters processing algorithm is based on the Bidirectional Encoder Representations from Transformers (BERT) algorithm and is used to determine whether the deviation between the measured values of the instrument and the theoretical calculated values exceeds the set threshold. The anomaly identification algorithm of nuclear power system is based on Sparse Auto Encoder (SAE) and Isolation Forest (iForest) algorithm to judge the operating state of nuclear power system. When the transient operating parameters processing algorithm judges that the deviation values have exceeded the set threshold, but the anomaly identification algorithm judges that the nuclear power system is in normal operation, it can determine that the current alarm signal is the false alarm. The false alarm detection algorithm of nuclear power system can not only provide judgment basis for the operators to analyze the state of nuclear power system, but also improve the intelligent level of nuclear power system, so as to further improve the safety and reliability of nuclear power system. [ABSTRACT FROM AUTHOR]

In the field of traditional industrial control, anomaly detection method is mainly used to identify data items that do not match the normal operation state of the system. The traditional machine learning algorithm needs the transient operation data of normal and accident conditions to identify the abnormal state of nuclear power plant. The transient operation data of nuclear power plant during normal condition is sufficient, but it is lacks of transient operation data in accident conditions. To solve the above problems, an abnormal operation state detection method of nuclear power plant based on unsupervised deep generative model is established by using Variational Auto Encoders (VAE) and Isolation Forest (iForest) in this paper. The biggest advantage of this method is that it can only make use of the normal operation data of the nuclear power plant to make the nuclear power plant control system effectively identify whether the current state of nuclear power plant is normal operation or in accident condition. In the unsupervised deep generative model, VAE is used for data preprocessing, and iForest is used to identify abnormal operation data. Then, the method is verified in seven variable and accident conditions, such as power reduction condition, steam generator tube rupture accident and so on. The verification results show that the anomaly detection method can recognize the current abnormal condition immediately when the accident happens. And the time consumed to identify a group of operation parameters corresponding to the operation state of the nuclear power plant is about 3 ms, which can satisfy the real-time requirements of the control system. Therefore, the anomaly detection method based on unsupervised deep generative model can distinguish the normal or abnormal operation state of the nuclear power plant in real time and effectively, and provide judgment basis for accident classification and subsequent rescue. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents a novel ranking framework for content-based multimedia information retrieval (CBMIR). The framework introduces relevance features and a new ranking scheme. Each relevance feature measures the relevance of an instance with respect to a profile of the targeted multimedia database. We show that the task of CBMIR can be done more effectively using the relevance features than the original features. Furthermore, additional performance gain is achieved by incorporating our new ranking scheme which modifies instance rankings based on the weighted average of relevance feature values. Experiments on image and music databases validate the efficacy and efficiency of the proposed framework. [Copyright &y& Elsevier]

Estimating the quality of GNSS vectors is decisive in planning GNSS networks and in several land surveying activities. These vectors are indirectly present in many civil infrastructures usually in the first stages of constructions and thus of high importance. In this research we assembled over 1000 baselines producing an extensive database with over 170,000 processed vectors. We propose a novel identification of outlying GNSS vectors based on the Isolation Forest (IF) algorithm based on the vectors deviations. And a new procedure to build linear models based on metaheuristics and a penalty function. The linear regressions presented models with a coefficient of determination R 2 up to 0.996. The observation time span variable remained in all equations at least twice, showing its importance for the outcome quality of a vector. Overall, the three-dimensional deviation of vectors processed with broadcast ephemeris is 2.4 times higher than for precise ephemerides. • Identification of outlying GNSS vectors based on the Isolation Forest algorithm. • GNSS vectors' models for deviation estimation with precise and broadcast ephemeris. • New procedure for models building combining metaheuristic and penalty function. [ABSTRACT FROM AUTHOR]

• This method adopts unsupervised learning and does not need to label the experimental data. • The method proposed in this paper is more suitable for data structures in real life. • A winding fault detection method of synchronous machine based on isolated forest and IFRA is proposed, which overcomes the shortcomings of traditional methods, such as long time-consuming, low accuracy, demand for massive fault experimental data. Synchronous machine is one of the critical power generation parts in the power system. Its stable operation ensures people's normal economic activities. Winding is an essential component of a synchronous machine, and the winding fault is a common fault type. The reliable and efficient fault diagnosis of synchronous machine winding is of great significance to ensure the stability of the power system. Therefore, this paper proposes an anomaly detection method of synchronous machine winding fault based on isolation forest (IF) and impulse frequency response analysis (IFRA). Firstly, the basic principle of the anomaly detection method is introduced, and mathematical-statistical indicators of IFRA signatures used are then explained. Besides, the experimental verification is carried out on a 5 kW synchronous machine, and the performance of the anomaly detection method for winding fault is compared with other conventional methods. The experimental results show that the proposed method is feasible and effective, and the generalization ability of the data is strong. The comparative experimental results show that the proposed method is superior to the existing conventional supervised learning method. It has a shorter calculation time and higher accuracy, with stronger robustness, which is more suitable for the actual data structure. [ABSTRACT FROM AUTHOR]

• Monitoring of mining conveyor using distributed optical fibre sensor is reported. • Optical signal is used to classify the various type of fault and its progression. • System architecture for automatic large data analysis and management is developed. • Automatic fault detection is proposed by modifying iForest algorithm. Condition monitoring of mining conveyor is a highly essential task to ensure minimum disruption to the mining operational system. Failure of one or more conveyor components can result in significant operational downtime, economic loss, and safety risks. The current monitoring method still involves subjective measure from maintenance engineers, where at some cases, fault can be left undetected and leads into site incident. Therefore, there is a high demand for real-time condition monitoring technology to detect early fault on conveyor. In this study, the effective application of distributed optical fibre sensor (DOFS) was explored for long distance real-time condition monitoring of mining conveyor. The fault detection framework was developed by integrating and modifying the Isolation Forest algorithm to analyse optical signals for effective detection of defective idlers. Further, the optical signal was analysed to extract the damage progression of defective idler with time and space. The results were used to classify various levels of damage and to set appropriate damage thresholds. Also, software interface, that can be used to set the sensing parameters, to collect, analyse, and visualise the signal in real-time, was developed. Finally, the developed condition monitoring system was used to monitor a 1.6 km long section of a conveyor structure in Western Australia for a period of 10 months. The results and findings from the field monitoring were presented together with automated fault detection framework for condition monitoring of mining conveyor. [ABSTRACT FROM AUTHOR]

Machine learning techniques provide useful methods for high-dimensional geochemical anomaly detection for mineral exploration targeting. However, the instability of the machine learning models often leads to the uncertainty of high-dimensional geochemical anomaly detection result. Combining various individual models to form an adaptive ensemble anomaly detector is a feasible way to enhance the robustness of machine learning anomaly detectors. In this study, the average method, maximization method, average of maximum (AOM) method, and maximum of average (MOA) method were adopted to combine the outputs of various k -nearest neighbor (KNN) anomaly detectors to improve the robustness of the KNN models in the high-dimensional geochemical anomaly detection in the Baishan district (Jilin Province, China). The effectiveness of the four combination algorithms for high-dimensional geochemical anomaly detection was evaluated by comparing the ensemble models obtained by using the four combination algorithms with the single KNN model, Gaussian mixture model (GMM), one-class support vector machine (OCSVM), and isolation forest (IForest) in the case study. It is found that the four ensemble models (a) perform similarly well in high-dimensional geochemical anomaly detection, and (b) perform better than the single KNN model, GMM, OCSVM, and IForest in high-dimensional geochemical anomaly detection. Therefore, the average method, maximization method, AOM method, and MOA method are potentially useful algorithms for combining the outputs of various KNN models to form robust ensemble models for high-dimensional geochemical anomaly detection. • A KNN model is used to detect high-dimensional geochemical anomalies. • Robust high-dimensional geochemical anomaly detection models are formed by combining the outputs of various KNN models. • The robust models are compared with KNN, GMM, OCSVM and IForest. [ABSTRACT FROM AUTHOR]

• We propose novel anomaly scores for Isolation Forests. • We design weighted variants of the anomaly score that embed additional information. • We propose a novel aggregation function to combine the scores at forest level. • Experiments confirm the suitability of the approach. • The novel proposed scores outperform the original one. Isolation Forest represents a variant of Random Forest largely and successfully employed for outlier detection. The main idea is that outliers are likely to get isolated in a tree after few splits. The anomaly score is therefore a function inversely related to the leaf depth. This paper proposes enhanced anomaly scores of the Isolation Forest by making two different contributions. The first consists in weighing the path traversed by an object to obtain a more informative anomaly score. The second contribution employs a different aggregation function to combine the tree scores. We thoroughly evaluate the proposed methodology by testing it on sixteen datasets. [ABSTRACT FROM AUTHOR]