Search

• Newbie study using standard ML techniques with HAR Application and Discussions. • Activities found in Literature with the corresponding reference. • Co-occurrences between activities and sensors with the corresponding reference. • Summary and comparison among the different datasets found in Literature. • Summary of the experimentation settings with performance scores found in Literature. Human Activity Recognition (HAR) is an essential area of research related to the ability of smartphones to retrieve information through embedded sensors and recognize the activity that humans are performing. Researchers have recognized people's activities by processing the data received from the sensors with Machine Learning Models. This work is intended to be a hands-on survey with practical's tables capable of guiding the reader through the sensors used in modern smartphones and highly cited developed machine learning models that perform human activity recognition. Several papers in the literature have been studied, paying attention to the preprocessing, feature extraction, feature selection, and classification techniques of the HAR system. In addition, several summary tables illustrating HAR approaches have been provided: most popular human activities in the literature with paper references, the most popular datasets available for download (Analyzing their characteristics, such as the number of subjects involved, the activities recorded, and the sensors with online-availability), co-occurrences between activities and sensors, and a summary table showing the performance obtained by researchers. =The paper's goal is to recommend, through the discussion phase and thanks to the tables, the current state of the art on this topic. [ABSTRACT FROM AUTHOR]

• Pinch force strength is accurately predicted using sEMG. • Different pinch movements are accurately classified using sEMG. • Similarity of EMG features is evaluated. • An EMG feature-to-image method is proposed. The movement of intelligent EMG-driven prosthesis mainly relies on the synergy of different fingers to achieve function of grasping objects. The paper proposes a novel scheme for force prediction and movement classification about pinch between different fingers based on surface electromyography (sEMG) using machine learning. The pinch force and sEMG signals are recorded synchronously by a data acquisition device. Eight features are extracted, which are proven to have better performance in the estimation of sEMG-to-force. We present a novel feature selection method that uses the one-dimensional time series similarity assessment based on Manhattan distance to eliminate the repetitive information between features. Three optimal features carrying less repetitive information are retained. Seven machine learning algorithms are used to predict force strength. The results show that the Long Short Term Memory (LSTM) has the best performance of force prediction, achieving a R 2 of 0.9517 and RMSE of 3.2723. The paper proposes a novel method of converting EMG feature sequences to the normalized gray image in order to classify the finger movement. Five classifiers based on image feature extraction and the Convolutional Neural Network (CNN) are developed respectively. The experimental results indicate that the CNN performs best, achieving an accuracy of 97.66%. In this way, it not only realizes the accurate force prediction, but also realizes the movement classification between different fingers. The proposed methodology has the potential to realize simultaneous force and movement control of prosthetic hand. [ABSTRACT FROM AUTHOR]

Identifying the abnormal samples from a data set and determining their type are two key tasks of anomaly detection. However, the existing anomaly detection algorithms are generally faced with the defects of weak generalization ability and insufficient interpretation, the core reason of which is that they cannot mine specific features for different abnormal types. In this paper, a new anomaly detection algorithm aiming at feature selection for different abnormal types is developed. Inspired by self-supervised learning, we take the stationarity of variance changes of abnormal score similarity as a pretext task and combine it with wrapped search method. Then, the features and the corresponding parameters for different abnormal types can be screened to apply to the downstream task of anomaly integration detection. To verify the efficiency of the new algorithm, we conduct two sets of experiments to compare the new algorithm with 11 classical anomaly detection and 3 clustering anomaly detection algorithms on the data sets WDBC, WPBC and Wilt from DAMI database with the evaluation measures P@n , Adj- P@n , AP, Adj-AP and AUC. The experiment results show that, both in the identification and classification on abnormal samples, all performance measures of the new algorithm are explicitly better than that of the contrastive algorithms. Also, we apply the new algorithm to the Chinese auto insurance market, and find that the results can help managers to identify the main patterns of fraudulent claims and to summarize the feature combinations of fraud behaviors. In general, the new algorithm developed in this paper has the following advantages compared with traditional algorithms: 1) It can directly capture abnormal features and realize effective recognition of abnormal types, which effectively bridge the gap between abnormal judgement and feature screening. 2) Automatic screening of abnormal features can be completed under the condition of self-updating learning optimal strategy. 3) Only a few features are extracted from all features to reveal the abnormal characteristics, which significantly improves the interpretability and generalization ability of the algorithm and its results. In a word, through the novel self-supervised design method, feature screening is skillfully integrated into the anomaly detection task, which may provide a new way for anomaly detection research. [ABSTRACT FROM AUTHOR]

Highlights • Extracting features using syntactic rules from pros/cons reviews. • Determining the opinion's polarity according to their emotion strength. • Comparing results with state-of-the-art approaches in different tasks. Abstract Sentiment knowledge extraction is a growing area of research in the literature. It helps in analyzing users’ opinions about different entities or events, which can then be utilized by analysts for various purposes. Particularly, feature-based sentiment analysis is one of the challenging research areas that analyzes users’ opinions on various features of a product or service. Of the three formats for the product reviews, our focus in this paper is limited to analyzing the pros/cons type. Due to the nature of pros/cons reviews, they are mostly concise and follow a different structure from other review types. Therefore, specialized techniques are needed to analyze these reviews and extract the customers’ discussed product features along with their personal attitudes. In this paper, we propose the Pros/Cons Sentiment Analyzer (PCSA) framework that exploits dependency relations in extracting sentiment knowledge from pros/cons reviews. We also utilize two different lexicons to ascertain the polarity strength of the extracted features based on the customers’ opinions. Several experiments are conducted to evaluate the performance of PCSA in its different phases. [ABSTRACT FROM AUTHOR]

Due to the presence of redundant and irrelevant features in large-dimensional biomedical datasets, the prediction accuracy of disease diagnosis can often be decreased. Therefore, it is important to adopt feature extraction methodologies that can deal with problem structures and identify underlying data patterns. In this paper, we propose a novel approach called the Anti Coronavirus Optimized Kernel-based Softplus Extreme Learning Machine (ACO-KSELM) to accurately predict different types of skin cancer by analyzing high-dimensional datasets. To evaluate the proposed ACO-KSELM method, we used four different skin cancer image datasets: ISIC 2016, ACS, HAM10000, and PAD-UFES-20. These dermoscopic image datasets were preprocessed using Gaussian filters to remove noise and artifacts, and relevant features based on color, texture, and shape were extracted using color histogram, Haralick texture, and Hu moment extraction approaches, respectively. Finally, the proposed ACO-KSELM method accurately predicted and classified the extracted features into Basal Cell Carcinoma (BCC), Squamous Cell Carcinoma (SCC), Actinic Keratosis (ACK), Seborrheic Keratosis (SEK), Bowen's disease (BOD), Melanoma (MEL), and Nevus (NEV) categories. The analytical results showed that the proposed method achieved a higher rate of prediction accuracy of about 98.9%, 98.7%, 98.6%, and 97.9% for the ISIC 2016, ACS, HAM10000, and PAD-UFES-20 datasets, respectively. [ABSTRACT FROM AUTHOR]

• 3D-CNNpred is the first 3-dimensional CNN model designed for stock market prediction. • CNNpred successfully combines various sources of information for prediction. • CNNs filters are designed to better handle financial data. • Deep CNN-based framework significantly outperforms shallow ANNs. • CNNpred is profitable in 4 out of 5 tested indices in presence of transaction costs. Feature extraction from financial data is one of the most important problems in market prediction domain for which many approaches have been suggested. Among other modern tools, convolutional neural networks (CNN) have recently been applied for automatic feature selection and market prediction. However, in experiments reported so far, less attention has been paid to the correlation among different markets as a possible source of information for extracting features. In this paper, we suggest a CNN-based framework, that can be applied on a collection of data from a variety of sources, including different markets, in order to extract features for predicting the future of those markets. The suggested framework has been applied for predicting the next day's direction of movement for the indices of S&P 500, NASDAQ, DJI, NYSE, and RUSSELL based on various sets of initial variables. The evaluations show a significant improvement in prediction's performance compared to the state of the art baseline algorithms. [ABSTRACT FROM AUTHOR]

Sensor-based human activity recognition (HAR) has drawn extensive attention from the research community due to its potential applications in various domains, including interactive gaming, activity monitoring, healthcare, etc. Although plentiful approaches (i.e., handcrafted feature-based and deep learning methods) have been proposed throughout the years, there are still several challenges in developing an efficient and effective HAR system. For instance, handcrafted feature-based methods rely on manual feature engineering by experts and require time-consuming feature selection methods. Conversely, deep learning methods can automatically capture salient features without domain experts. However, some deep learning methods, especially Convolutional Neural Networks (CNN), cannot extract temporal features effectively, which are significant to motion analysis. Unlike CNN, recurrent models are exceptional at capturing temporal characteristics, but these models contain gigantic model parameters, requiring tremendous computation. This may limit the deployment of such models, especially to low-spec or embedded devices. Hence, this paper proposes a lightweight deep learning model, Lightweight Multiheaded TCN (Light-MHTCN), for human activity recognition. Light-MHTCN extracts the multiscale features of the inertial sensor signals through the parallelly organised Convolutional Heads to capture richer information. Further, integrating dilated causal convolutions and residual connections preserves longer-term dependency, which can boost the overall model performance. The performance of Light-MHTCN is assessed on three popular smartphone-based HAR databases: UCI HAR, WISDM V1 and UniMiB SHAR. With only ∼ 0.21 million parameters, our lightweight model is able to achieve state-of-the-art performance with recognition accuracies of 96.47%, 99.98% and 98.63% on these databases, respectively. • Light-MHTCN requires minimal preprocessing and no manual feature engineering. • Light-MHTCN is lightweight in computation, with only 0.21M parameters. • Light-MHTCN allows multiscale feature extraction due to the parallel architecture. • Light-MHTCN retains longer-term dependency using dilations and residual connections. • Light-MHTCN achieves 96.47% on UCI HAR, 99.98% on WISDM V1 and 98.63% on UniMiB SHAR. [ABSTRACT FROM AUTHOR]

The quality monitoring of the laser shock peening (LSP) process is of great significance for improving the intelligence of precision manufacturing. At present, the monitoring method combining acoustic emission (AE) and machine learning inevitably faces the problems of large data samples, lack of artificial feature experience, large parameter redundancy and slow recognition speed. To overcome these limitations, this paper proposes a new method of LSP monitoring combining AE time-scale texture images and multi-scale blueprint separable convolutional networks with attention mechanism (MBSCN-AM). Firstly, this method discards the tedious and lossy manual feature extraction process, and integrates the time scale information into the texture image. Secondly, a feature extraction module of multi-scale convolution and a feature selection module of the attention mechanism are designed using advanced blueprint separable convolution as the underlying architecture. The multi-scale convolution kernel optimizes the balance between details and overall information through different receptive fields. The adjustment of high-level features is made possible by the attention mechanism, which evaluates and adaptively assigns weights. To verify the validity of the proposed method, LSP experiments were designed to collect AE signals at six different laser energies. The experimental results show that the proposed method can effectively distinguish between different stress states on the material surface compared to other existing monitoring methods. On the premise of faster calculation and lower storage requirements, the proposed method has higher recognition accuracy and stability. Furthermore, the operation mechanism of the proposed method is elaborated through ablation experiments and visual analysis. [ABSTRACT FROM AUTHOR]

Highlights • Aggregate acoustic and visual features for acoustic scene classification. • Investigate three feature selection methods for acoustic scene classification. • Compare time-frequency representations for visual feature extraction. Abstract Acoustic scene classification has gained great interests in recent years due to its diverse applications. Various acoustic and visual features have been proposed and evaluated. However, few studies have investigated acoustic and visual feature aggregation for acoustic scene classification. In this paper, we investigated various feature sets based on the fusion of acoustic and visual features. Specifically, acoustic features are directly extracted from the waveform: spectral centroid, spectral entropy, spectral flux, spectral roll-off, short-time energy, zero-crossing rate, and Mel-frequency Cepstral coefficients. For visual features, we calculate local binary pattern, histogram of gradients, and moments based on the audio scene time-frequency representation. Then, three feature selection algorithms are applied to various feature sets to reduce feature dimensionality: correlation-based feature selection, principal component analysis, and ReliefF. Experimental results show that our proposed system was able to achieve an accuracy improvement of 15.43% compared to the baseline system with the development set. When all development sets are used for training, the performance based on the evaluation set provided by the TUT Acoustic scene 2016 challenge is 87.44%, which is the fourth best among all non-neural network systems. [ABSTRACT FROM AUTHOR]

Highlights • Features are extracted from raw bridge acceleration data from 8 bridges to provide network training data; no FE models are necessary for training. • Multiple neural network voting ensemble configurations are presented. • Impact detection ranges from 91 −100% while average false positive rates are 0.00–0.75%. Abstract Many critical societal functions depend on uninterrupted service of civil engineering infrastructure. Railroads represent important infrastructure components of the transportation sector and provide both passenger and freight services. Railroad bridges over roadways are susceptible to impacts from overheight vehicles and equipment, which may damage bridge girders or supports and must be investigated after each event. One method of monitoring for vehicle-bridge collisions utilizes accelerometers to monitor for abnormal bridge vibrations corresponding to abnormal activity. Passing trains under normal operating conditions frequently produce significant bridge responses that have similar response characteristics to bridge strikes, but do not need to be investigated. This paper presents an expert system which comprises committees of artificial neural networks trained to interrogate data collected from accelerometers mounted on the bridge, assess the nature of the acceleration signal, and classify the event as either a passing train or a potentially damaging impact. This system is trained using acceleration time histories from accelerometers installed on 8 low-clearance rail bridges; no finite element model simulations were used for network training or data stream creation. The presented system accurately detects and classifies impacts with average impact detection performance ranging from 91–100% with average false positive rates limited to 0.00–0.75%. [ABSTRACT FROM AUTHOR]

Highlights • Grasshopper optimization algorithm is used to propose a feature selection method. • Computation of feature goodness factor helps to find global solution. • The accuracy of 100% is achieved by selecting 4 features out of 2308 in a dataset. • The method achieves the highest classification accuracy in 7 out of 10 datasets. Abstract Feature selection is the problem of finding the minimum number of features among a redundant feature space which leads to the maximum classification performance. In this paper, we have proposed a novel feature selection method based on mathematical model of interaction between grasshoppers in finding food sources. Some modifications were applied to the grasshopper optimization algorithm (GOA) to make it suitable for a feature selection problem. The method, abbreviated as GOFS is supplemented by statistical measures during iterations to replace the duplicate features with the most promising features. Several publicly available datasets with various dimensionalities, number of instances, and target classes were considered to evaluate the performance of the GOFS algorithm. The results of implementing twelve well-known and recent feature selection methods were presented and compared with GOFS algorithm. Comparative experiments indicate the significance of the proposed method in comparison with other feature selection methods. [ABSTRACT FROM AUTHOR]

Highlights • We define and address the rare category exploration problem with noisy labels. • We propose a compactness based similarity matrix for RCE. • Experimental evaluation demonstrates the effectiveness of our two methods. • We prove the effectiveness of our methods via theoretical analysis. Abstract Starting from a few labelled data examples as the seeds, rare category exploration (RCE) aims to find out the target rare category hidden in the given dataset. However, the performance of conventional RCE approaches is very sensitive to noisy labels while the presence of noises in manually generated labels is almost inevitable. To address this deficiency of traditional RCE approaches, this paper investigates the RCE process in the presence of noisy labels, which to the best of our knowledge has not yet been intensively studied by previous research. Based on the assumption that only one labelled data example of the rare category is correctly labelled while the other few data examples may be wrongly labelled, we first propose a label propagation based algorithm SLP to extract the coarse shape of a rare category. Then, we refine the result by proposing a mixture-information based propagation model, RLP. Extensive experiments have been conducted on six real-world datasets, which show that our method outperforms the state-of-the-art RCE approaches. We also show that even with 20% noisy labels, our method is able to achieve a satisfactory accuracy. [ABSTRACT FROM AUTHOR]

Highlights • Convexity is used as a cue for fruit recognition in RGB-D images. • A novel convex template instance (CTI) descriptor is used for fruit recognition. • A dataset of four sorts of fruits acquired in an unstructured environment is provided. • Nearest neighbor approach is used for classification. • Advantages of the CTI descriptor are low dimensionality and fast computation. Abstract Automatic fruit picking is a challenging problem in robotics with a wide application field. A prerequisite for realization of a robotic fruit picker is its ability to detect fruits in tree tops. An expert system, which would be able to compete with human perception, must be capable of recognizing fruits among leaves and branches under uncontrolled conditions, where fruits are occluded and shaded. In this paper, a novel approach for fruit recognition in RGB-D images based on detection and classification of convex surfaces is proposed. The input RGB-D image is first segmented into convex surfaces by a region growing procedure. Each convex surface is then described by an appropriate descriptor and classified with the aid of the associated descriptor. A novel descriptor of approximately convex surfaces is proposed, which we named Convex Template Instance (CTI) descriptor. It is based on approximating surfaces by convex polyhedrons with quantized face orientations, where every polyhedron face corresponds to one descriptor component. Computation of the proposed descriptor is simple and can be performed very efficiently. The proposed CTI descriptor is compared to the SHOT descriptor, a standard descriptor for 3D point clouds. Two variants of the both CTI and SHOT descriptor are evaluated, a variant which uses color and a variant which does not. A k-nearest neighbor classifier is used to classify detected surfaces into two classes: fruit and other. The main advantage of the proposed expert system in comparison to other fruit recognition solutions is its computational efficiency, which is of great importance for its target application – an automatic fruit picker. The proposed approach is evaluated using a challenging dataset containing RGB-D images of four fruit sorts acquired under uncontrolled conditions, which has been made publicly available to the scientific community, allowing benchmarking of novel fruit recognition methods. [ABSTRACT FROM AUTHOR]

Highlights • We propose a new joint neural model for entity recognition and relation extraction. • Our model does not rely on external NLP tools nor hand-crafted features. • Entities and relations within the same sentence are extracted simultaneously. • We experimentally evaluate extensively on four datasets (ACE04, CoNLL04, ADE, DREC). • Our model achieves state-of-the-art performance in different contexts and languages. Abstract State-of-the-art models for joint entity recognition and relation extraction strongly rely on external natural language processing (NLP) tools such as POS (part-of-speech) taggers and dependency parsers. Thus, the performance of such joint models depends on the quality of the features obtained from these NLP tools. However, these features are not always accurate for various languages and contexts. In this paper, we propose a joint neural model which performs entity recognition and relation extraction simultaneously, without the need of any manually extracted features or the use of any external tool. Specifically, we model the entity recognition task using a CRF (Conditional Random Fields) layer and the relation extraction task as a multi-head selection problem (i.e., potentially identify multiple relations for each entity). We present an extensive experimental setup, to demonstrate the effectiveness of our method using datasets from various contexts (i.e., news, biomedical, real estate) and languages (i.e., English, Dutch). Our model outperforms the previous neural models that use automatically extracted features, while it performs within a reasonable margin of feature-based neural models, or even beats them. [ABSTRACT FROM AUTHOR]

Highlights • Features that discriminate classes are linked to provide an undirected graph. • Features relationships are defined based on unsupervised subspace learning. • PageRank is investigated to rank features according to their importance in graph. • High dimension low sample data are used to assess and compare the proposed method. • The proposed unsupervised graph based method outperforms competitive methods. Abstract Feature selection has become an indispensable part of intelligent systems, especially with the proliferation of high dimensional data. It identifies the subset of discriminative features leading to better learning performances, i.e., higher learning accuracy, lower computational cost and significant model interpretability. This paper proposes a new efficient unsupervised feature selection method based on graph centrality and subspace learning called UGFS for ‘ Unsupervised Graph-based Feature Selection ’. The method maps features on an affinity graph where the relationships (edges) between feature nodes are defined by means of data points subspace preference. Feature importance score is then computed on the entire graph using a centrality measure. For this purpose, we investigated the Google’s PageRank method originally introduced to rank web-pages. The proposed feature selection method has been evaluated using classification and redundancy rates measured on the selected feature subsets. Comparisons with the well-known unsupervised feature selection methods, on gene/expression benchmark datasets, demonstrate the validity and the efficiency of the proposed method. [ABSTRACT FROM AUTHOR]

Highlights • We Illustrated weaknesses of balanced accuracy and normalized difference measures. • We proposed a new feature ranking metric called max-min ratio (MMR). • MMR better estimates the true worth of a term in high class skews. • We tested MMR against 8 well-known metrics on 6 datasets with 2 classifiers. • MMR statistically outperforms metrics in 76% macro F1 cases and 74% micro F1 cases. Abstract Text classification automatically assigns text documents to one or more predefined categories based on their content. In text classification, data are characterized by a large number of highly sparse terms and highly skewed categories. Working with all the terms in the data has an adverse impact on the accuracy and efficiency of text classification tasks. A feature selection algorithm helps in selecting the most relevant terms. In this paper, we propose a new feature ranking metric called max-min ratio (MMR). It is a product of max-min ratio of the true positives and false positives and their difference, which allows MMR to select smaller subsets of more relevant terms even in the presence of highly skewed classes. This results in performing text classification with higher accuracy and more efficiency. To investigate the effectiveness of our newly proposed metric, we compare its performance against eight metrics (balanced accuracy measure, information gain, chi-squared, Poisson ratio, Gini index, odds ratio, distinguishing feature selector, and normalized difference measure) on six data sets namely WebACE (WAP, K1a, K1b), Reuters (RE0, RE1), and 20 Newsgroups using the multinomial naive Bayes (MNB) and support vector machines (SVM) classifiers. The statistical significance of MMR has been estimated on 5 different splits of training and test data sets using the one-way analysis of variance (ANOVA) method and a multiple comparisons test based on Tukey–Kramer method. We found that performance of MMR is statistically significant than that of the other 8 metrics in 76.2% cases in terms of macro F 1 measure and in 74.4% cases in terms of micro F 1 measure. [ABSTRACT FROM AUTHOR]

F-score is a simple feature selection technique, however, it works only for two classes. This paper proposes a novel feature ranking method based on Fisher discriminate analysis (FDA) and F-score, denoted as FDAF-score, which considers the relative distribution of classes in a multi-dimensional feature space. The main idea is that a proper subset is got according to maximizing the proportion of average between-class distance to the relative within-class scatter. Because the method removes all insignificant features at a time, it can effectively reduce computational cost. Experiments on six benchmarking UCI datasets and two artificial datasets demonstrate that the proposed FDAF-score algorithm can not only obtain good results with fewer features than the original datasets as well as fast computation but also deal with the classification problem with noises well. [ABSTRACT FROM AUTHOR]

The curse of dimensionality is a common problem in classification tasks. However, feature selection is an exciting approach to deal with this type of problem by searching for a suboptimal feature set, either by eliminating irrelevant attributes, those with redundant information, or even both. Furthermore, although the use of machine learning techniques for the classification of drum cymbals is found in the literature, little attention has been paid to approaches that focus on classifying these instruments according to their bronze alloys. This paper aims to explore and evaluate the temporal information retrieved from audios, using TSFEL (Time Series Feature Extraction Library) as a tool to extract 18 temporal attributes, three feature selection approaches to assess these features, and logistic regression as a classifier. To this end, 276 audios referring to four drum cymbals of three different bronze alloys were captured in the studio through a sound acquisition procedure that took into account environment and microphone variations. Hence, one expects to find an optimal subset of features that contains enough information from the audios to achieve the best classification performance in the proposed problem. The experimental results show that a feature selection approach sequentially based on L 1 Regularization and Correlation Analysis was able to find a subset consisting of 5 attributes that achieved an average accuracy of 97.08%. • The computational model can identify the cymbals' bronze alloys through their sound. • A standardized impact procedure is developed for the audio signals acquisition. • Feature selection could reduce the dimensionality up to 70%, improving accuracy. • Logistic Regression with feature selection achieved high material classification rate. • High accuracy is attained even considering different sound capture perspectives. [ABSTRACT FROM AUTHOR]

The goal of few-shot learning is to use limited labeled samples to complete independent classification tasks. The feature extractor of few-shot learning needs to have a stronger feature expression ability to generalize in unseen novel classes. To further enhance the expressive ability, in this paper, we propose an inherited feature extraction method, named Base and Meta Feature Extraction (BMFE). Base feature represents the task-irrelevant classification information of each sample. Meta feature obtained by the proposed Triplet Meta-train Mechanism (TMM) inherits the classification information and also contains the task-related meta information of each sample. We concatenate both the base and meta features to complementarily express the rich information of each sample. Besides, instead of relying on limited support samples to obtain the prototype, we propose a novel unsupervised prototype correction module, named Prototype Self-updating (PSU). All unlabeled query samples in a few-shot test task participate in the iterative updating of each prototype in the task without training. Extensive experiments prove that our overall method can obtain richer features by BMFE and more accurate prototypes by PSU. Our overall method outperforms state-of-the-art methods on miniImageNet and tiredImageNet datasets, and especially under the 1-shot case we obtains 78.45% and 81.21% classification accuracy respectively. • BMFE can obtain the complementary features representation of FSL. • TMM mechanism can better perform meta-training of BMFE. • PSU can correct prototypes without training process. • The proposed CFPSU gets outstanding results on two standard datasets. [ABSTRACT FROM AUTHOR]

Disease diagnosis and prediction methods in biotechnology and medicine have significantly advanced over time. Consequently, analyzing raw gene expression is crucial for identifying diseases such as cancer. Interestingly, microarrays are a tool that records gene expression from deoxyribonucleic acid (DNA) or ribonucleic acid. This technique exhibits intriguing characteristics, such as generating high-dimensional data with a small sample size. However, in the case of such dataset, the classification model is prone to overfitting. This limitation can be overcome by reducing the dimensions of the microarray datasets to a reasonable number. Machine learning (ML)-based data reduction has recently achieved considerable attention in genomic research. Therefore, this review examines recent studies that present state-of-the-art data reduction and classification algorithms for microarray gene expression data to diagnose tumors and analyzes their performance. To the best of our knowledge, this is the first review that provides a comprehensive view of data preprocessing, dimensionality reduction, including feature (i.e., gene) selection, feature extraction, and their hybrid, and ML algorithms. The paper is structured as follows. First, this review summarizes several data preprocessing methods applied to gene expression datasets. Then, a detailed review of various ML-based feature selection algorithms, including filter, wrapper, embedded, ensemble, and hybrid algorithms, is discussed. These algorithms are examined under three main classes—supervised, unsupervised, and semisupervised ML. Next, the feature extraction and hybrid of feature extraction and selection algorithms are thoroughly reviewed. Furthermore, a detailed review of broadly applied ML algorithms to simplify tumor and nontumor classification using microarray datasets is presented. Finally, the challenges and open questions related to gene expression datasets for accurate cancer classification and detection are highlighted. [Display omitted] • A comprehensive review on microarray gene expression reduction is proposed. • A novel taxonomy of up-to-date gene reduction algorithms is presented. • The strength and limitations of each type of gene reduction algorithms are provided. • Provide an outline of the accurately used type of machine learning algorithms. • Analyze the performance of the state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

• This paper proposes a new system for ear biometric identification. • It selects the best combination of local and global spectral features using GA. • The global features are extracted by applying the Gabor-Zernike operator. • The local phase quantization operator is used for extracting local features. Identification systems based on biometric features are becoming increasingly important. One of the most common biometric features is the ear. The accuracy of these systems is heavily dependent on the characteristics extracted from them. In this paper, an appropriate combination of local and global features in the frequency domain is extracted as unique features of the ear region. In the proposed approach, at first the image quality is improved by Contrast-limited Adaptive Histogram Equalization. Then, the global features of the ear region are extracted by applying the Gabor-Zernike operator to the whole image and its non-overlapping blocks. In addition, to extract of local features, the local phase quantization operator is used on the original image of the ear region. Then, the optimum combination of global and local features is selected using Genetic Algorithm. Finally, the nearest neighbor classifier with Canberra distance is used to identify users. The proposed approach is evaluated using three databases, i.e. USTB-1, IIT125 and IIT221. The recognition rate of 100%, 99.2% and 97.13%, is reported on these databases, respectively. The obtained results show that the proposed approach performs better than existing ear recognition methods. [ABSTRACT FROM AUTHOR]

The selection of the best features to be used in expert systems is a key issue in obtaining a satisfactory performance. Unsupervised speaker segmentation and clustering is the task of the automatic identification of the number of participants in a meeting and the determination of their speaking turns (also called “diarization”). This is part of an intelligent system that replaces human intervention in several tasks related to automatic language and speech processing. The segmentation and clustering of speakers is crucial if we want to transcribe any audio recording automatically when several people take their turn. It is a task necessary to archive automatically interventions of several people in meetings, broadcast radio, lectures, parliamentary sessions etc. since a simple transcription of what is said without assigning it to a specific speaker makes the information unusable. The automation of this task would save enormous amounts of resources currently spent on human transcribers. When used online it could also be useful to point a video camera automatically to the person talking when a videoconference with multiple speakers is taking place thus replacing a human operator. Furthermore it could also help to scan large amounts of audio automatically in search of crimes or audio interventions of a particular person. In the case of recordings with several distant microphones (MDM), spatial features may and should be used. The most widely used spatial features in diarization are the Time Delay of Arrival (TDOA) features. These delays are extracted from pairs of microphones of unknown location and quality, which makes the selection of the best pairs highly advisable. This paper analyses this issue and proposes and evaluates several methods that significantly improve the performance both in speaker error rate (SER) and in computational time. The methods propose a selection ofTDOA features based on the quality of the cross-correlation of signals coming from different pairs of microphones. We prove that the use of the wrong pairs can be highly detrimental to the overall performance. The methods proposed, based on cross correlation, are compared and combined with other two selection methods, based on the dynamic range of the delay features and the selection of every pair of microphones available followed by a reduction in dimensionality. Although all algorithms achieve some improvements, it is proved that selection methods based on cross correlation have the fewest errors. The improvements on the baseline system for the two best proposed systems are 25.14% and 33.70% for the development set, and 55.06% and 46.09% for the test set. Furthermore the best method for the test set also reduces the computational cost by 20%. [ABSTRACT FROM AUTHOR]

• Remora Whale Optimization Based-Hybrid deep model is used for detecting intrusions. • RWO is designed by the combination of ROA and WOA. • In the proposed approach features are selected by holoentropy process. • The developed approach achieves superior performance with testing accuracy of 0.93. Security remains as a key role in this internet world owing to the fast expansion of users on the internet. Numerous existing intrusion detection approaches were introduced by numerous researchers to recognize and identify intruders. Meanwhile, the existing systems failed to achieve satisfactory detection accuracy. Hence, this paper develops a robust intrusion detection model, named Remora Whale Optimization (RWO)-based Hybrid deep model for detecting intrusions. Here, the input data is pre-processed, and thereafter data transformation is done. With the transformed data, effective CNN features are extracted and feature conversion is performed to convert the features into vector form. Moreover, RV-coefficient is accomplished for performing feature selection process and finally, network intrusions are effectively detected using Hybrid deep model where the Deep Maxout Network and Deep Auto Encoder are used. On the other hand, the training procedure of the Hybrid deep model is carried out using the designed optimization algorithm, named RWO, which is the hybridization of the Remora Optimization Algorithm (ROA) and Whale Optimization Algorithm (WOA). Furthermore, the devised technique achieved superior performance using the evaluation metrics, such as testing accuracy, precision, recall, and F1-score with the higher values of 0.938, 0.920, 0.932, and 0.926, respectively. [ABSTRACT FROM AUTHOR]

• 3 layer neural networks outperform other ML algorithms in automated assessments. • In subjective assignments, deep neural networks can improve score prediction. • Word embeddings and linguistic feature results in more precise text mining models. • Extraction and selection of appropriate features can improve results. Machine learning (ML) has recently gained popularity in a variety of domains for automating tasks. This has also been used in the evaluation of student responses/ assignments. This paper presents a qualitatively enhanced methodology for automated score prediction of subjective assignments. To consider all the major aspects of a human grader, 21 linguistic features related to syntactical, grammatical, sentimental, and readability are analyzed from the assignments. This study makes use of the ASAP competition dataset from Kaggle. The evaluation metric used is Quadratic Weighted Kappa (QWK), which measures the agreement between the human graded score and the predicted score. The effect of appropriate feature selection has been observed using Mutual Information Regression. Four ML algorithms are investigated on identified linguistic features. 3 Layer Neural Network with feature selection performed well among all chosen ML algorithms with average QWK of 0.678. To include the benefits of deep learning, a new hybrid model (LF-BiLSTM-att-FS) is proposed that combines a higher level deep neural network (DNN) with the selected features. Pre-trained Glove embedding is used to include contextual information of the text. The proposed model results demonstrated an improvement in overall accuracy, with an average QWK value of 0.768. [ABSTRACT FROM AUTHOR]

• Application of uncapacitated P-median problem to obtain clustered tweets. • Applying a novel hybrid genetic bat algorithm on uncapacitated P-median problem. • Generation of a similarity matrix via Latent Dirichlet Allocation model. • Determining needs of the victims after an earthquake using Twitter data. Due to the rapid incline in the number of documents along with social media usage, text categorization has become an important concept. There are tasks required to be fulfilled during the text categorization, such as extracting useful data from different perspectives, reducing the high feature space dimension, and improving effectiveness. In order to accomplish these tasks, feature selection, and feature extraction gain importance. This paper investigates how to solve feature selection and extraction problems. Also, this study aims to decide which topics are the focus of a document. Moreover, the Twitter data-set is utilized as a document and an Uncapacitated P-Median Problem (UPMP) is applied to make clustering. In this study, UPMP is used on Twitter data collection for the first time to collect clustered tweets. Therefore, a novel hybrid genetic bat algorithm (HGBA) is proposed to solve the UPMP for our case. The proposed novel approach is applied to analyze the Twitter data-set of the Nepal earthquake. The first part of the analysis includes the data pre-processing stage. The Latent Dirichlet Allocation (LDA) method is applied to the pre-processed text. After that, a similarity (distance) matrix is generated by utilizing the Jensen Shannon Divergence (JSD) model. The study's main goal is to use Twitter to assess the needs of victims during and after a disaster. To evaluate the applicability of the proposed approach, experiments are conducted on the OR-Library data-set. The results demonstrate that the proposed approach successfully extracts topics and categorizes text. [ABSTRACT FROM AUTHOR]

• Several operations are handled for bi-traits-centered offender identification. • Image quality is highly increased by the proposed method of offender identification. • The system solves the computation issues of tree structure part model. • A new way is used for weight value selection in DNN. • To prove the effectiveness of the proposed algorithms. The evidence, traits, in addition to clues that are amassed as of the Crime Scenes (CS) are analyzed by a process termed Forensic application. Nevertheless, the Forensic Analysis (FA) faces the challenges of irrelevant Feature Extraction (FE), which is because of the over impression of Finger-Print (FP), facial expressions, pose-invariant, together with poor lightings that bring about the mis-identification of the offender. To trounce such challenges, this paper proposes a bi-traits-based offender identification using Cross Entropy-based Deep Remainder Neural Network (CEDRNN) for forensic application. Pre-processing, FE, Feature Selection (FS), together with identification are '4′ phases that the proposed model encompasses. In the pre-processing phase, the input data is processed via modifying the image size, augmenting the Image Contrast (IC), and neglecting the unwanted data. Then, as of the pre-processed data, the more apt and helpful data are extracted in FE. Then, the Gaussian Sailfish Optimization (GSO) technique selects the important data in the FS phase. Finally, the CEDRNN identifies the offender centered on the selected features. In this research, publically available datasets are utilized for comparing the outcomes of the CEDRNN with the previous top-notch algorithms. The experimental outcomes exhibited that the CEDRNN attains the highest accuracy and also effectively identifies the offender without mis-prediction. [ABSTRACT FROM AUTHOR]

Urban and road planners must take right decisions related to urban traffic management and controlling noise pollution. Their assessments and resolutions have important consequences on the annoyance of population exposed to road-traffic-noise and controlling other environmental pollutants (e.g. NOx or ultrafine particles emitted by heavy vehicles). One of the key decisions is the selection of which noise control actions should be taken in sensitive areas (residential or hospital areas, school areas etc), that could include costly measures such as reducing the overall traffic, banning or reducing traffic of heavy vehicles, inspection of motorbikes sound emission, etc. For an efficient decision-making in noise control actions, it is critical to classify a given location in a sensitive area according to the different prevailing traffic conditions. This paper outlines an expert system aimed to help urban planners to classify urban locations based on their traffic composition. To induce knowledge into the system, several machine learning algorithms are used, based on multi-layer Perceptron and support vector machines with sequential minimal optimization. As input variables for these algorithms, a combination of environment variables was used. For the development of the classification models, four feature selection techniques, i.e., two subset evaluation (correlation-based feature-subset selection and consistency-based subset evaluation) and two attribute evaluation (ReliefF and minimum redundancy maximum relevance) were implemented to reduce the models’ complexity. The overall procedure was tested on a full database collected in the city of Granada (Spain), which includes urban locations with road-traffic as dominant noise source. Among all the possibilities tested, support vector machines based models achieves the better results in classifying the considered urban locations into the 4 categories observed, with values of average weighted F -measure and Kappa statistics (used as indicators) up to 0.9 and 0.8. Regarding the feature selection techniques, attribute evaluation algorithms (ReliefF and mRMR) achieve better classification results than subset evaluation algorithms in reducing the model complexity, and so relevant environmental variables are chosen for the proposed procedure. Results show that these tools can be used for addressing a prompt assessment of potential road-traffic-noise related problems, as well as for gathering information in order to take more well-founded actions against urban road-traffic noise. [ABSTRACT FROM AUTHOR]

This paper proposes hybrid feature selection approaches based on the Genetic Algorithm (GA). This approach uses a hybrid search technique that combines the advantages of filter feature selection methods with an enhanced GA (EGA) in a wrapper approach to handle the high dimensionality of the feature space and improve categorization performance simultaneously. First, we propose EGA by improving the crossover and mutation operators. The crossover operation is performed based on chromosome (feature subset) partitioning with term and document frequencies of chromosome entries (features), while the mutation is performed based on the classifier performance of the original parents and feature importance. Thus, the crossover and mutation operations are performed based on useful information instead of using probability and random selection. Second, we incorporate six well-known filter feature selection methods with the EGA to create hybrid feature selection approaches. In the hybrid approach, the EGA is applied to several feature subsets of different sizes, which are ranked in decreasing order based on their importance, and dimension reduction is carried out. The EGA operations are applied to the most important features that had the higher ranks. The effectiveness of the proposed approach is evaluated by using naïve Bayes and associative classification on three different collections of Arabic text datasets. The experimental results show the superiority of EGA over GA, comparisons of GA with EGA showed that the latter achieved better results in terms of dimensionality reduction, time and categorization performance. Furthermore, six proposed hybrid FS approaches consisting of a filter method and the EGA are applied to various feature subsets. The results showed that these hybrid approaches are more effective than single filter methods for dimensionality reduction because they were able to produce a higher reduction rate without loss of categorization precision in most situations. [ABSTRACT FROM AUTHOR]

Feature extraction methods for sound events have been traditionally based on parametric representations specifically developed for speech signals, such as the well-known Mel Frequency Cepstrum Coefficients (MFCC). However, the discrimination capabilities of these features for Acoustic Event Classification (AEC) tasks could be enhanced by taking into account the spectro-temporal structure of acoustic event signals. In this paper, a new front-end for AEC which incorporates this specific information is proposed. It consists of two different stages: short-time feature extraction and temporal feature integration. The first module aims at providing a better spectral representation of the different acoustic events on a frame-by-frame basis, by means of the automatic selection of the optimal set of frequency bands from which cepstral-like features are extracted. The second stage is designed for capturing the most relevant temporal information in the short-time features, through the application of Non-Negative Matrix Factorization (NMF) on their periodograms computed over long audio segments. The whole front-end has been evaluated in clean and noisy conditions. Experiments show that the removal of certain frequency bands (which are mainly located in the medium region of the spectrum for clean conditions and in low frequencies for noisy environments) in the short-time feature computation process in conjunction with the NMF technique for temporal feature integration improves significantly the performance of a Support Vector Machine (SVM) based AEC system with respect to the use of conventional MFCCs. [ABSTRACT FROM AUTHOR]

Falls in elderly people are becoming an increasing healthcare problem, since life expectancy and the number of elderly people who live alone have increased over recent decades. If fall detection systems could be installed easily and economically in homes, telecare could be provided to alleviate this problem. In this paper we propose a low cost fall detection system based on a single wide-angle camera. Wide-angle cameras are used to reduce the number of cameras required for monitoring large areas. Using a calibrated video system, two new features based on the gravity vector are introduced for fall detection. These features are: angle between the gravity vector and the line from feet to head of the human and size of the upper body. Additionally, to differentiate between fall events and controlled lying down events the speed of changes in the features is also measured. Our experiments demonstrate that our system is 97% accurate for fall detection. [ABSTRACT FROM AUTHOR]

• A stacked reinforced discriminant autoencoder is proposed for feature extraction. • The proposed stacked autoencoder and MRMR are combined for feature selection. • A stacked supervised t-SNE is proposed for data visualization. • A new visualization-based process monitoring method is introduced. Visual process monitoring is to monitor industrial processes by projecting the high-dimensional process data into the two-dimensional space, which provides powerful insight for industrial processes, and accelerates fault diagnosis. The challenge of visual process monitoring lies in how to project the complex process data into the two-dimensional plane and separate different classes as much as possible. In this paper, a new visual process monitoring method is proposed. First, a stacked reinforced discriminant auto-encoder (SRDAE) which consists of multiple reinforced discriminant auto-encoders (RDAEs) is proposed to extract discriminant features. In SRDAE, the useful features in the original data and the hidden output of the previous RDAE are combined together as the input of the latter RDAE, and the error of class label is added into the loss function of RDAE. Therefore, SRDAE can prevent the loss of useful information in the original data in the high layers and make the extracted features have the powerful ability to separate different classes. Furthermore, in order to extract the more informative discriminant features, minimal redundancy maximal relevance (MRMR) technology is utilized to select important neurons from all layers of the SRDAE as the final feature representation of the original data. Finally, a stacked supervised t-distributed stochastic neighbor embedding network is proposed to visualize the discriminant features for process monitoring. The effectiveness of the proposed method is validated on the Tennessee Eastman process, the experiments show that the proposed method can effectively separate different classes to achieve intuitive and efficient process monitoring. [ABSTRACT FROM AUTHOR]

Feature selection in real-world data mining problems is essential to make the learning task efficient and more accurate. Identifying the best feature selection algorithm, among the many available, is a complex activity that still relies heavily on human experts or some random trial-and-error procedure. Thus, the automated machine learning community has taken some steps towards the automation of this process. In this paper, we address the metalearning challenge of recommending feature selection algorithms by proposing a novel meta-feature engineering model. Our model considers a broad collection of meta-features that enable the study of the relationship between the dataset properties and the feature selection algorithm performance in terms of several criteria. We arrange the input meta-features into eight categories: (i) simple, (ii) statistical, (iii) information-theoretical, (iv) complexity, (v) landmarking, (vi) based on symbolic models, (vii) based on images, and (viii) based on complex networks (graphs). The target meta-features emerge from a multi-criteria performance measure, based on five individual performance indexes, that assesses feature selection methods grounded in information, distance, dependence, consistency, and precision measures. We evaluate our proposal using a recently developed framework that extracts the input meta-features from 213 benchmark datasets, and ranks the assessed feature selection algorithms, to fill in the target meta-features in meta-bases. This evaluation uses five state-of-the-art classification methods to induce recommendation models from meta-bases: C4.5, Random Forest, XGBoost, ANN, and SVM. The results showed that it is possible to reach an average accuracy of up to 90% applying our meta-feature engineering model. This work is the first to use an extensive empirical evaluation to provide a careful discussion of the strengths and limitations of more than 160 meta-features. These meta-features, while designed to aid the task of feature selection algorithm recommendation, can readily be employed in other metalearning scenarios. Therefore, we believe our findings are a valuable contribution to the fields of automated machine learning and data mining, as well as to the feature extraction and pattern recognition communities. • A novel meta-feature engineering model recommends feature selection algorithms. • The proposal obtains promising results from 213 datasets with hit rates of up to 90%. • Some simple, landmarking, image, and graph-based input meta-features highlighted. • A multi-criteria performance measure rigorously assesses candidate algorithms. • Chains of binary or multiclass classifiers can efficiently rank candidate algorithms. [ABSTRACT FROM AUTHOR]

• The study proposes a novel method of using inertial sensors for game play analysis. • Data from accelerometer and gyroscope is used for expert-novice classification. • Early and late fusion of features is evaluated along with wrapper feature selection. • State-of-the-art classification accuracy is achieved using smartphone inertial data. The gaming industry has seen a tremendous growth in the last decade due to an exponential increase in the number of smartphone users. Embedded smartphone sensors provide solutions for automatic game controls during game-play. In this paper, we present an experimental study for the expertise classification of a game (mobile-based) player using smartphone inertial sensors, while they are simultaneously used for game controls. The game expertise level of participants is either labeled as expert or novice using game scores. Towards this end, data from 38 participants are curated during Traffic Racer game-play (in three different trials) using the embedded gyroscope and accelerometer sensors of the smartphone. These signals are pre-processed using Savitzky-Golay smoothing filter to remove noise. Twenty time domain features are extracted from the pre-processed data and are subjected to the wrapper-based feature selection method to select an optimum subset of features. Three classifiers, including k-nearest neighbor (k-NN), random forest, and the Naive Bayes, are evaluated towards the classification of player's expertise level, i.e., expert and novice. The best average accuracy of 92.1 % is achieved with k-NN classifier using the fusion of gyroscope and accelerometer data, which outperforms the existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

• A novel feature ranking technique for bearing fault diagnosis is proposed. • Several machine learning and artificial intelligence classifiers are used for validation. • Fewer feature subset is required with proposed methodology. • It has higher classification accuracy with less time consumption. Bearing failure can cause hazardous effects on rotating machinery. The diagnosis of the fault is very critical for reliable operation. The main steps for the machine learning process involve feature extraction, selection, and classification. Feature selection contains an identification of noble features that performs for better classification accuracy with fewer features and with less computational time. For a large feature dimension; a critical study is required to catch the best feature subset for proper diagnosis. So, this paper presents a unique feature ordering and selection technique called Feature Ranking and Subset Selection based on Euclidean distance (FRSSED). Two bearing databases have considered for verification of the robustness of the proposed technique. One database was obtained from the experiment, and the other publicly available database was collected from Case Western Reserve University (CWRU). Initially, the vibration signals have captured from bearings having an individual as well as combined defects in various components along with healthy bearing. EEMD was applied to these signals, and then, the sensitive IMF was selected by the envelope spectrum. In the later stage, the feature extraction was carried out from the selected IMF using fifteen statistical features. Afterward, the extracted features were introduced into FRSSED algorithm for feature ordering. These ordered features were fed into various classifiers. The comparison was made for classification accuracy and time consumption among generalized method (without feature ordering), principal component analysis (PCA), and FRSSED. The diagnostic outcomes describe that the suggested feature reduction technique improves the classification accuracy with fewer feature subset along with considerable time-saving. [ABSTRACT FROM AUTHOR]

• A fast hybrid dimensionality reduction method for classification is proposed. • Multi-strategy based feature selection is used to filter out irrelevant features. • Grouped feature extraction is used to remove redundancy among features. • The proposed method shows excellent efficiency and competitive classification performance. Dimensionality reduction is one basic and critical technology for data mining, especially in current "big data" era. As two different types of methods, feature selection and feature extraction each have their pros and cons. In this paper, we combine multi-strategy feature selection and grouped feature extraction and propose a novel fast hybrid dimension reduction method, incorporating their advantages of removing irrelevant and redundant information. Firstly, the intrinsic dimensionality of the data set is estimated by the maximum likelihood estimation method. Fisher Score and Information Gain based feature selection are used as multi-strategy methods to remove irrelevant features. With the redundancy among the selected features as clustering criterion, they are grouped into a certain amount of clusters. In every cluster, Principal Component Analysis (PCA) based feature extraction is carried out to remove redundant information. Four classical classifiers and representation entropy are used to evaluate the classification performance and information loss of the reduced set. The runtime results of different methods show that the proposed hybrid method is consistently much faster than the other three in almost all of the sets used. Meanwhile, the proposed method shows competitive classification performance, which has no significant difference basically compared with the other methods. The proposed method reduces the dimensionality of the raw data fast and it has excellent efficiency and competitive classification performance compared with the contrastive methods. [ABSTRACT FROM AUTHOR]