Search

Timely detection and prevention of violations in the technological process of wastewater treatment caused by threats of different nature is a highly relevant research problem. Modern systems are equipped with a large number of technological sensors. Data from these sensors can be used to detect anomalies in the technological process. Their timely detection, prediction and processing ensures the continuity and fault tolerance of the technological process. The aim of the research is to improve the accuracy of detection of such anomalies. We propose a methodology for the identification and subsequent assessment of cyber resilience risks of the wastewater treatment process, which includes the distinctive procedure of training dataset generation and the anomaly detection based on deep learning methods. The availability of training datasets is a necessary condition for the efficient application of the proposed technology. A distinctive feature of the anomaly detection approach is a new method of processing input sensor data, which allows the use of computationally efficient analytical models with high accuracy of anomaly detection, and outperforms the efficiency of previously published methods.

Understanding the intricate pathways of protein secretion in bacteria is crucial for advancing research on bacterial diseases and their potential treatments, particularly in the case of non-classical protein secretion pathways. These pathways pose unique challenges due to the complex signaling mechanisms involved. To address this, we employed advanced machine learning techniques and gathered physical–chemical characteristics of amino acids from the AA index site. Through a meticulous six-step methodology, we curated a comprehensive dataset by filtering raw genome data and juxtaposing it with a positive dataset comprising 141 proteins from authoritative literature sources. Leveraging a conventional Random Forest machine learning algorithm, we achieved an impressive accuracy rate of approximately 91% in classifying non-classical secreted proteins. This validation was conducted on a dataset of 14 positive and 92 negative proteins, resulting in a sensitivity of 91% and a specificity of 86%. Notably, our study distinguishes itself by its rapid execution of non-classical secretion pathway analyses, rendering it particularly suitable for large datasets. This speed does not compromise accuracy, allowing for comprehensive Omics analyses. Consequently, our research underscores the significance of carefully selecting appropriate descriptors and constructing a robust training dataset to enhance Omics analyses of bacterial protein secretion via non-classical pathways. For further details, please refer to the complete study available at https://github.com/santosardr/non-CSPs. [ABSTRACT FROM AUTHOR]

This study addresses a significant gap in the field of time series regression modeling by highlighting the central role of data augmentation in improving model accuracy. The primary objective is to present a detailed methodology for systematic sampling of training datasets through data augmentation to improve the accuracy of time series regression models. Therefore, different augmentation techniques are compared to evaluate their impact on model accuracy across different datasets and model architectures. In addition, this research highlights the need for a standardized approach to creating training datasets using multiple augmentation methods. The lack of a clear framework hinders the easy integration of data augmentation into time series regression pipelines. Our systematic methodology promotes model accuracy while providing a robust foundation for practitioners to seamlessly integrate data augmentation into their modeling practices. The effectiveness of our approach is demonstrated using process data from two milling machines. Experiments show that the optimized training dataset improves the generalization ability of machine learning models in 86.67% of the evaluated scenarios. However, the prediction accuracy of models trained on a sufficient dataset remains largely unaffected. Based on these results, sophisticated sampling strategies such as Quadratic Weighting of multiple augmentation approaches may be beneficial.

This study addresses a significant gap in the field of time series regression modeling by highlighting the central role of data augmentation in improving model accuracy. The primary objective is to present a detailed methodology for systematic sampling of training datasets through data augmentation to improve the accuracy of time series regression models. Therefore, different augmentation techniques are compared to evaluate their impact on model accuracy across different datasets and model architectures. In addition, this research highlights the need for a standardized approach to creating training datasets using multiple augmentation methods. The lack of a clear framework hinders the easy integration of data augmentation into time series regression pipelines. Our systematic methodology promotes model accuracy while providing a robust foundation for practitioners to seamlessly integrate data augmentation into their modeling practices. The effectiveness of our approach is demonstrated using process data from two milling machines. Experiments show that the optimized training dataset improves the generalization ability of machine learning models in 86.67% of the evaluated scenarios. However, the prediction accuracy of models trained on a sufficient dataset remains largely unaffected. Based on these results, sophisticated sampling strategies such as Quadratic Weighting of multiple augmentation approaches may be beneficial. [ABSTRACT FROM AUTHOR]

We describe 20 datasets derived through signal filtering and feature extraction steps applied to the raw time series EEG data of 20 epileptic patients, as well as the methods we used to derive them. Background: Epilepsy is a complex neurological disorder which has seizures as its hallmark. Electroencephalography plays a crucial role in epilepsy assessment, offering insights into the brain's electrical activity and advancing our understanding of seizures. The availability of tagged training sets covering all seizure phases—inter-ictal, pre-ictal, ictal, and post-ictal—is crucial for data-driven epilepsy analyses. Methods: Using the sliding window technique with a two-second window length and a one-second time slip, we extract multiple features from the preprocessed EEG time series of 20 patients from the Freiburg Seizure Prediction Database. In addition, we assign a class label to each instance to specify its corresponding seizure phase. All these operations are made through a software application we developed, which is named Training Builder. Results: The 20 tagged training datasets each contain 1080 univariate and bivariate features, and are openly and publicly available. Conclusions: The datasets support the training of data-driven models for seizure detection, prediction, and clustering, based on features engineering. Dataset: Data are available at https://doi.org/10.5281/zenodo.10808054. Dataset License: CC-BY 4.0 [ABSTRACT FROM AUTHOR]

Current research in artificial intelligence (AI) sheds light on algorithmic bias embedded in AI systems. The underrepresentation of women in the AI design sector of the tech industry, as well as in training datasets, results in technological products that encode gender bias, reinforce stereotypes and reproduce normative notions of gender and femininity. Biased behaviour is notably reflected in anthropomorphic AI systems, such as personal intelligent assistants (PIAs) and chatbots, that are usually feminized through various design parameters, such as names, voices and traits. Gendering of AI entities, however, is often reduced to the encoding of stereotypical behavioural patterns that perpetuate normative assumptions about the role of women in society. The impact of this behaviour on social life increases, as human-to-(anthropomorphic)machine interactions are mirrored in human-to-human social interactions. This article presents current critical research on AI bias, focusing on anthropomorphic systems. Moreover, it discusses the significance of women's engagement in AI design and programming, by presenting selected case studies of contemporary female media artists and designers. Finally, it suggests that women, through their creative practice, provide feminist and critical approaches to AI design which are essential for imagining alternative, inclusive, ethic and de-biased futures for anthropomorphic AIs. [ABSTRACT FROM AUTHOR]

With the advancement of deep learning (DL), researchers and engineers in the marine industry are exploring the application of DL technologies to their specific applications. In general, the accuracy of inference using DL technologies is significantly dependent on the number of training datasets. Unfortunately, people in marine science and engineering environments are often reluctant to share their documents (i.e., P&ID) with third-party manufacturers or public clouds to protect their proprietary information. Despite this, the demand for object detection using DL technologies in image-formatted files (i.e., jpg, png, or pdf format) is steadily growing. In this paper, we propose a new mechanism, called a no-training object picker (NoOP), which efficiently recognizes all objects (e.g., lines, tags, and symbols) in image-formatted P&ID documents. Notably, it can recognize objects without any training dataset, thus reducing the time and effort required for training and collection of unpublished datasets. To clearly present the effectiveness of NoOP, we evaluated NoOP using a real P&ID document. As a result, we confirmed that all objects in the image-formatted P&ID file are successfully detected over a short time (only 7.11 s on average). [ABSTRACT FROM AUTHOR]

We describe 20 datasets derived through signal filtering and feature extraction steps applied to the raw time series EEG data of 20 epileptic patients, as well as the methods we used to derive them. Background: Epilepsy is a complex neurological disorder which has seizures as its hallmark. Electroencephalography plays a crucial role in epilepsy assessment, offering insights into the brain’s electrical activity and advancing our understanding of seizures. The availability of tagged training sets covering all seizure phases—inter-ictal, pre-ictal, ictal, and post-ictal—is crucial for data-driven epilepsy analyses. Methods: Using the sliding window technique with a two-second window length and a one-second time slip, we extract multiple features from the preprocessed EEG time series of 20 patients from the Freiburg Seizure Prediction Database. In addition, we assign a class label to each instance to specify its corresponding seizure phase. All these operations are made through a software application we developed, which is named Training Builder. Results: The 20 tagged training datasets each contain 1080 univariate and bivariate features, and are openly and publicly available. Conclusions: The datasets support the training of data-driven models for seizure detection, prediction, and clustering, based on features engineering.

In this paper, we present the Academic Results Datasets Predictor (ARDP), for missing academic results datasets, based on chi-squared expected calculation, positional clustering, progressive approximation of relative residuals, and positional averages of the data in a sampled population. Academic results datasets are data originating from inside academic institutions’ results repositories. It is a technique designed specifically for predicting missing academic results. Since the whole essence of data mining is to elicit useful information and gain knowledge-driven insights into datasets, ARDP positions data explorer at this advantageous perspective. ARDP is committed to solve missing academic results dataset problems more quickly over and above what currently obtains. PARD is computed by leveraging on the averages of neighbouring values. The predictor was implemented using Python, and the results show that it is admissible in a minimum of up to 85 percent accurate predictions of the sampled cases. It has been verified that ARDP shows a tendency toward greater precision in providing the best solution to the problems of predictions of missing academic results datasets in universities.

With the advancement of deep learning (DL), researchers and engineers in the marine industry are exploring the application of DL technologies to their specific applications. In general, the accuracy of inference using DL technologies is significantly dependent on the number of training datasets. Unfortunately, people in marine science and engineering environments are often reluctant to share their documents (i.e., P&ID) with third-party manufacturers or public clouds to protect their proprietary information. Despite this, the demand for object detection using DL technologies in image-formatted files (i.e., jpg, png, or pdf format) is steadily growing. In this paper, we propose a new mechanism, called a no-training object picker (NoOP), which efficiently recognizes all objects (e.g., lines, tags, and symbols) in image-formatted P&ID documents. Notably, it can recognize objects without any training dataset, thus reducing the time and effort required for training and collection of unpublished datasets. To clearly present the effectiveness of NoOP, we evaluated NoOP using a real P&ID document. As a result, we confirmed that all objects in the image-formatted P&ID file are successfully detected over a short time (only 7.11 s on average).

With the recent trends in urban agriculture and climate change, there is an emerging need for alternative plant culture techniques where dependence on soil can be eliminated. Hydroponic and aquaponic growth techniques have proven to be viable alternatives, but the lack of efficient and optimal practices for irrigation and nutrient supply limits its applications on a large-scale commercial basis. The main purpose of this research was to develop statistical methods and Machine Learning algorithms to regulate nutrient concentrations in aquaponic irrigation water based on plant needs, for achieving optimal plant growth and promoting broader adoption of aquaponic culture on a commercial scale. One of the key challenges to developing these algorithms is the sparsity of data which requires the use of Bolstered error estimation approaches. In this paper, several linear and non-linear algorithms trained on relatively small datasets using Bolstered error estimation techniques were evaluated, for selecting the best method in making decisions regarding the regulation of nutrients in hydroponic environments. After repeated tests on the dataset, it was decided that Semi-Bolstered Resubstitution Error estimation technique works best in our case using Linear Support Vector Machine as the classifier with the value of penalty parameter set to one. A set of recommended rules have been prescribed as a Decision Support System, using the output of the Machine Learning algorithm, which have been tested against the results of the baseline model. Further, the positive impact of the recommended nutrient concentrationson plant growth in aquaponic environments has been elaborately discussed.

The dynamic development of deep learning methods in recent years has prompted the widespread application of these algorithms in the field of photogrammetry and remote sensing, especially in the areas of image recognition, classification, and object detection. Still, one of the biggest challenges in this field is the low availability of training datasets, especially regarding applications of oblique aerial imagery and UAV data. The process of acquiring such databases is labor-intensive. The solution to the problem of the unavailability of datasets and the need for manual annotation is to automate the process of generating annotations for images. One such approach is used in the following work. The proposed methodology for semi-automating the creation of training datasets was applied to detect objects on nadir and oblique images acquired from UAV. The methodology includes the following steps: (1) the generation of a dense 3D point cloud by two different methods: UAV photogrammetry and TLS (terrestrial laser scanning); (2) data processing, including clipping to objects and filtering of point clouds; (3) the projection of cloud points onto aerial images; and (4) the generation of bounding boxes bounding the objects of interest. In addition, the experiments performed are designed to test the accuracy and quality of the training datasets acquired in the proposed way. The effect of the accuracy of the point cloud extracted from dense UAV image matching on the resulting bounding boxes extracted by the proposed method was evaluated. [ABSTRACT FROM AUTHOR]

The purpose of this study was to construct artificial intelligence (AI) training datasets based on multi-resolution remote sensing and analyze the results through learning algorithms in an attempt to apply machine learning efficiently to (quasi) real-time changing landcover data. Multi-resolution datasets of landcover at 0.51- and 10-m resolution were constructed from aerial and satellite images obtained from the Sentinel-2 mission. Aerial image data (a total of 49,700 data sets) and satellite image data (300 data sets) were constructed to achieve 50,000 multi-resolution datasets. In addition, raw data were compiled as metadata in JavaScript Objection Notation format for use as reference material. To minimize data errors, a two-step verification process was performed consisting of data refinement and data annotation to improve the quality of the machine learning datasets. SegNet, U-Net, and DeeplabV3+ algorithms were applied to the datasets; the results showed accuracy levels of 71.5%, 77.8%, and 76.3% for aerial image datasets and 88.4%, 91.4%, and 85.8% for satellite image datasets, respectively. Of the landcover categories, the forest category had the highest accuracy. The landcover datasets for AI training constructed in this study provide a helpful reference in the field of landcover classification and change detection using AI. Specifically, the datasets for AI training are applicable to large-scale landcover studies, including those targeting the entirety of Korea.

The dynamic development of deep learning methods in recent years has prompted the widespread application of these algorithms in the field of photogrammetry and remote sensing, especially in the areas of image recognition, classification, and object detection. Still, one of the biggest challenges in this field is the low availability of training datasets, especially regarding applications of oblique aerial imagery and UAV data. The process of acquiring such databases is labor-intensive. The solution to the problem of the unavailability of datasets and the need for manual annotation is to automate the process of generating annotations for images. One such approach is used in the following work. The proposed methodology for semi-automating the creation of training datasets was applied to detect objects on nadir and oblique images acquired from UAV. The methodology includes the following steps: (1) the generation of a dense 3D point cloud by two different methods: UAV photogrammetry and TLS (terrestrial laser scanning); (2) data processing, including clipping to objects and filtering of point clouds; (3) the projection of cloud points onto aerial images; and (4) the generation of bounding boxes bounding the objects of interest. In addition, the experiments performed are designed to test the accuracy and quality of the training datasets acquired in the proposed way. The effect of the accuracy of the point cloud extracted from dense UAV image matching on the resulting bounding boxes extracted by the proposed method was evaluated.

The capability of accurately predicting the Solar Photovoltaic (PV) power productions is crucial to effectively control and manage the electrical grid. In this regard, the objective of this work is to propose an efficient Artificial Neural Network (ANN) model in which 10 different learning algorithms (i.e., different in the way in which the adjustment on the ANN internal parameters is formulated to effectively map the inputs to the outputs) and 23 different training datasets (i.e., different combinations of the real-time weather variables and the PV power production data) are investigated for accurate 1 day-ahead power production predictions with short computational time. In particular, the correlations between different combinations of the historical wind speed, ambient temperature, global solar radiation, PV power productions, and the time stamp of the year are examined for developing an efficient solar PV power production prediction model. The investigation is carried out on a 231 kWac grid-connected solar PV system located in Jordan. An ANN that receives in input the whole historical weather variables and PV power productions, and the time stamp of the year accompanied with Levenberg-Marquardt (LM) learning algorithm is found to provide the most accurate predictions with less computational efforts. Specifically, an enhancement reaches up to 15, 1, and 5% for the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R2) performance metrics, respectively, compared to the Persistence prediction model of literature.

See also the commentary by Sitek in this issue. Supplemental material is available for this article.

Precision and effectiveness of Artificial Intelligence (AI) models are highly dependent on the availability of genuine, relevant, and representative training data. AI systems tested and validated on poor-quality datasets can produce inaccurate, erroneous, skewed, or harmful outcomes (actions, behaviors, or decisions), with far-reaching effects on individuals' rights and freedoms. Appropriate data governance for AI development poses manifold regulatory challenges, especially regarding personal data protection. An area of concern is compliance with rules for lawful collection and processing of personal data, which implies, inter alia, that using databases for AI design and development should be based on a clear and precise legal ground: the prior consent of the data subject or another specific valid legal basis. Faced with this challenge, the European Union's personal data protection legal framework does not provide a preferred, one-size-fits-all answer, and the best option will depend on the circumstances of each case. Although there is no hierarchy among the different legal bases for data processing, in doubtful cases, consent is generally understood by data controllers as a preferred or default choice for lawful data processing. Notwithstanding this perception, obtaining data subjects' consent is not without drawbacks for AI developers or AI-data controllers, as they must meet (and demonstrate) various requirements for the validity of consent. As a result, data subjects' consent could not be a suitable and realistic option to serve AI development purposes. In view of this, it is necessary to explore the possibility of basing this type of personal data processing on lawful grounds other than the data subject's consent, specifically, the legitimate interest of the data controller or third parties. Given its features, legitimate interests could help to meet the challenge of quality, quantity, and relevance of data curation for AI training. The aim of this article is to provide an initial conceptual approach to support the debate about data governance for AI development in the European Union (EU), as well as in non-EU jurisdictions with European-like data protection laws. Based on the rules set by the EU General Data Protection Regulation (GDPR), this paper starts by referring to the relevance of adequate data curation and processing for designing trustworthy AI systems, followed by a legal analysis and conceptualization of some difficulties data controllers face for lawful processing of personal data. After reflecting on the legal standards for obtaining data subject's valid consent, the paper argues that legitimate interests (if certain criteria are met) may better match the purpose of building AI training datasets. [ABSTRACT FROM AUTHOR]