Selection Approach Research Articles

Regularly Truncated M-Estimators for Learning With Noisy Labels.

The sample selection approach is very popular in learning with noisy labels. As deep networks "learn pattern first", prior methods built on sample selection share a similar training procedure: the small-loss examples can be regarded as clean examples and used for helping generalization, while the large-loss examples are treated as mislabeled ones and excluded from network parameter updates. However, such a procedure is arguably debatable from two folds: (a) it does not consider the bad influence of noisy labels in selected small-loss examples; (b) it does not make good use of the discarded large-loss examples, which may be clean or have meaningful information for generalization. In this paper, we propose regularly truncated M-estimators (RTME) to address the above two issues simultaneously. Specifically, RTME can alternately switch modes between truncated M-estimators and original M-estimators. The former can adaptively select small-losses examples without knowing the noise rate and reduce the side-effects of noisy labels in them. The latter makes the possibly clean examples but with large losses involved to help generalization. Theoretically, we demonstrate that our strategies are label-noise-tolerant. Empirically, comprehensive experimental results show that our method can outperform multiple baselines and is robust to broad noise types and levels.

Early prediction of end-stage kidney disease using electronic health record data: a machine learning approach with a 2-year horizon.

In the United States, end-stage kidney disease (ESKD) is responsible for high mortality and significant healthcare costs, with the number of cases sharply increasing in the past 2 decades. In this study, we aimed to reduce these impacts by developing an ESKD model for predicting its occurrence in a 2-year period. We developed a machine learning (ML) pipeline to test different models for the prediction of ESKD. The electronic health record was used to capture several kidney disease-related variables. Various imputation methods, feature selection, and sampling approaches were tested. We compared the performance of multiple ML models using area under the ROC curve (AUCROC), area under the Precision-Recall curve (PR-AUC), and Brier scores for discrimination, precision, and calibration, respectively. Explainability methods were applied to the final model. Our best model was a gradient-boosting machine with feature selection and imputation methods as additional components. The model exhibited an AUCROC of 0.97, a PR-AUC of 0.33, and a Brier score of 0.002 on a holdout test set. A chart review analysis by expert physicians indicated clinical utility. An ESKD prediction model can identify individuals at risk for ESKD and has been successfully deployed within our health system.

Performance evaluation of feature selections on some ML approaches for diagnosing the narcissistic personality disorder

Narcissistic personality disorder (NPD) is a personality disorder that affects various aspects of life, including relationships, employment, school, and finances. Persons with NPD usually feel unhappy and disappointed when no one helps them and is not praised for their achievements. Diagnosing narcissism is generally done using a screening test that consumes time and costs a lot. This research aims to evaluate the performance of several feature selection (FS) approaches on machine learning (ML) techniques (support vector machine (SVM), random forest classifier (RFC), and Naive Bayes). Three scenarios of FS (all features, the information gain technique and the gain ratio (GR) feature technique) are used for each ML method. Several experiments using the benchmark narcissistic disorder dataset have been done. It adopts the k-fold cross-validation (10-fold cross-validation) strategy. We evaluate the method’s performance by measuring its accuracy, error rate, and processing time. It is shown that the RFC GR strategy gives the best performance with an accuracy of 100%.

Theory-Guided Feature Selection in Cybercrime Data Science

Cybercrime data science is being significantly hampered by the presence of 'noisy' features within vast and complex datasets. We draw from the theoretical insights of the behavioural sciences to propose a feature selection model to enrich and improve the value and interpretability of cybercrime intelligence datasets. We piloted our theory-guided feature selection approach on a subset of intelligence datafeeds provided by a global fraud and cybercrime tracking firm. The results of the proposed social influence feature selection model show significant improvement in the interpretability of the machine learning-based exploratory analysis and advanced visualization techniques in an experimental setting. The feature selection model yielded rich insights about cybercriminal psychological tactics from social engineering scam data and has potential applicability in the areas of cyberthreat response and cybersecurity awareness training. Our study shows the value of an interdisciplinary theory-guided approach to cybercrime data analytics that integrates scientific knowledge from the behavioural sciences and data science expertise. Our paper concludes by suggesting avenues for future research on theory-guided feature selection seeking to incorporate behavioural science knowledge in cybercrime data science. We intend to refine, automate, evaluate, and scale our model in future research to assess its effectiveness in producing insights about cybercriminal activities and informing decision-making in a naturalistic and real-time setting. In future research efforts, we aim to automate the encoding of features and apply a wider range of machine learning tools and evaluation metrics to extract more meaningful insights into cybercriminal psychological tactics. We also intend to refine our model on larger datasets to enhance its efficiency and responsiveness to real-time cybercrime data.  We call on data scientists and cybercrime domain experts to work together to apply theory-guided feature selection to improve processes of knowledge discovery that enhance our cybersecurity capabilities.

LSFSR: Local label correlation-based sparse multilabel feature selection with feature redundancy

In recent studies, existing multilabel feature selection models have focused on either considering the relationship between labels or the redundancy between features. Furthermore, they only use simple sparsity constraints to process high-dimensional data without the intrinsic relationships between features and labels. These issues can have a great impact on the classification effectiveness of feature selection. To address these limitations, this article describes a new local label correlation-based sparse multilabel feature selection approach with feature redundancy. First, a new loss function is established among the matrices of samples, label coefficients, and labels. Then, the Frobenius norm is imposed to investigate the potential relationships between features and labels. The weight matrix is sparsified by the l2,1 norm to ensure that the new loss function has high interpretability. Second, a manifold constraint is employed to capture the local geometric structure between labels and to delve deeper into the latent information among the local labels. Then manifold constraints and Laplacian scores are combined for embedding feature selection to guide the exploration of hidden latent label. Finally, by considering the differences between the feature scores and the redundancy between the samples, feature redundancy is analyzed via the modified cosine similarity, and a candidate feature subset with low redundancy is generated. The l2 norm is used to select features with low redundancy while preserving sparsity, and a novel objective function is developed to optimize this solution. Thus, a sparse feature selection algorithm via local label correlation and feature redundancy is designed, and has demonstrated remarkable classification effectiveness in comparative experiments conducted on 21 multilabel datasets.

Assessment of Genetic Variability, Heritability, Genetic Advance, and Diversity Analysis for Morpho Physiological Traits in Finger Millet [Eleusine coracana (l.) Gaertn.]

The study, conducted during the Kharif season of 2022, aimed to evaluate genetic variability, heritability, genetic advance, and diversity for morpho-physiological traits in finger millet (Eleusine coracana (L.) Gaertn.) using 39  genotypes. Significant genetic variation was identified for all studied traits through variance analysis. While moderate to high values of Phenotypic Coefficient of Variation (PCV) and Genotypic Coefficient of Variation (GCV) were observed for most traits, exceptions were noted for days to 50 percent flowering, days to maturity, relative water content in leaves, and chlorophyll stability index. The study revealed high heritability with moderate to high genetic advance as a percent of the mean for most traits, suggesting the prevalence of additive gene effects. However, chlorophyll stability index and relative water content exhibited lower values. D2 analysis indicated the existence of sufficient genetic diversity among the genotypes. The genotypes were categorized into five distinct non-overlapping clusters.  Cluster I, the largest cluster with 35 genotypes, displayed maximum intra-cluster distance. Clusters V and III had the highest inter-cluster distances, Days to maturity played a significant role in overall divergence. The findings suggest that a direct selection approach in finger millet landraces could lead to a substantial increase in yield response. This research provides valuable insights into the genetic characteristics of finger millet genotypes, offering a foundation for targeted breeding programs to enhance crop improvement.

Comparison of Domain Selection Methods for Multi-Source Manifold Feature Transfer Learning in Electroencephalogram Classification

Transfer learning (TL) utilizes knowledge from the source domain (SD) to enhance the classification rate in the target domain (TD). It has been widely used to address the challenge of sessional and inter-subject variations in electroencephalogram (EEG)-based brain–computer interfaces (BCIs). However, utilizing knowledge from a combination of both related and non-related sources can significantly deteriorate the classification performance across individual target domains, resulting in a negative transfer (NT). Hence, NT becomes one of the most significant challenges for transfer learning algorithms. Notably, domain selection techniques have been developed to address the challenge of NT emerging from the transfer of knowledge from non-related sources. However, existing domain selection approaches iterate through domains and remove a single low-beneficial domain at a time, which can massively affect the classification performance in each iteration since SDs respond differently to other sources. In this paper, we compare domain selection techniques for a multi-source manifold feature transfer learning (MMFT) framework to address the challenge of NT and then evaluate the effect of beneficial and non-beneficial sources on TL performance. To evaluate the effect of low-beneficial and high beneficial sources on TL performance, some commonly used domain selection methods are compared, namely, domain transferability estimation (DTE), rank of domain (ROD), label similarity analysis, and enhanced multi-class MMFT (EMC-MMFT), using the same multi-class cross-session and cross-subject classification problems. The experimental results demonstrate the superiority of the EMC-MMFT algorithm in terms of minimizing the effect of NT. The highest classification accuracy (CA) of 100% is achieved when optimal combinations of high beneficial sources are selected for two-class SSMVEP problems, while the highest CAs of 98% and 87% are achieved for three- and four-class SSMVEP problems, respectively. The highest CA of 98% is achieved for two-class MI classification problems, while the highest CAs of 90% and 71.5% are obtained for three- and four-class MI problems, respectively.

Optimal Selection of Seed-Trees Using the Multi-Objective NSGA-II Algorithm and a Seed Dispersal Model

Optimal seed-tree selection during natural regeneration of shade-intolerant species requires ensuring an ample and uniform seed supply from residual trees with the smallest possible seed-tree density. Here, we propose a novel approach for seed-tree selection using the genetic algorithm. Data are derived from a 3-hectare even-aged stand of Pinus canariensis C.Sm. ex DC, comprising 364 mature trees and 103 seed-traps. Seeds were collected in 2007 and 2008. After constructing a seed-dispersal model for each seed-crop year, we employ the multi-objective non-dominated sorting genetic algorithm to identify the smallest seed-tree set that maximizes post-treatment seed supply and its spatial homogeneity. Optimal solutions range from a maximum of 68.4% to a minimum of 38.1% reduction in stand density, resulting in a 59.5% to 28% reduction in post-felling seed supply. The coefficient of variation of among-site seed-flux varies from 28% to 59.5%. Proposing a treatment involving the removal of 240 trees (65.9% stand-density reduction) and leaving 40 seed-trees per hectare, our findings provide insights into balancing the conflicting objectives of sufficient post-treatment seed supply at a minimum seed-tree density. This approach marks a departure from traditional practices, as the decision about which trees to cut is historically left to the discretion of field managers.

Current status of graft choice in adults for anterior cruciate ligament reconstruction: a systematic review and meta-analysis

The anterior cruciate ligament (ACL) is a frequently performed surgical procedure in the field of sports medicine, and graft selection is an essential aspect of ACL reconstruction (ACLR) that has been shown to optimize post-operative rehabilitation, facilitate return to full sporting function and reduce the risk of complications. However, there needs to be more agreement regarding optimal graft choice. The present study aims to identify the optimal graft choice in the adult population undergoing ACLR. The current systematic review explores the electronic database of online libraries of academic institutions from 2011 to 30 September 2022 and compared the clinical outcomes of autograft and allograft for ACLR, which provides an evidence-based approach for graft selection. The findings of our present study delineate that ACLR in adult populations with autograft is the most promising choice for ligament tear over allograft based on lower failure rate, reduced complications, and high cost-effectiveness. However, there was no significant difference in overall IKDC, pivot shift, Lachman test, and Tegner activity. This critical analysis recommends that for an adult population, ACLR with autograft can provide better clinical results than an allograft, but further studies carried out about other parameters affecting the long-term stability, quick return, and physical therapy adopted outside the supervision of the respective surgeon would be required.

Erratum to: A hybrid feature selection approach for urinary tract infection detection and prediction in IoT-fog environment

The online version of the original article can be found at https://doi.org/10.31893/multiscience.2024080

IBio-GATS-A Semi-Automated Workflow for Structural Modelling of Insect Odorant Receptors.

Insects utilize seven transmembrane (7TM) odorant receptor (iOR) proteins, with an inverted topology compared to G-protein coupled receptors (GPCRs), to detect chemical cues in the environment. For pest biocontrol, chemical attractants are used to trap insect pests. However, with the influx of invasive insect pests, novel odorants are urgently needed, specifically designed to match 3D iOR structures. Experimental structural determination of these membrane receptors remains challenging and only four experimental iOR structures from two evolutionarily distant organisms have been solved. Template-based modelling (TBM) is a complementary approach, to generate model structures, selecting templates based on sequence identity. As the iOR family is highly divergent, a different template selection approach than sequence identity is needed. Bio-GATS template selection for GPCRs, based on hydrophobicity correspondence, has been morphed into iBio-GATS, for template selection from available experimental iOR structures. This easy-to-use semi-automated workflow has been extended to generate high-quality models from any iOR sequence from the selected template, using Python and shell scripting. This workflow was successfully validated on Apocrypta bakeri Orco and Machilis hrabei OR5 structures. iBio-GATS models generated for the fruit fly iOR, OR59b and Orco, yielded functional ligand binding results concordant with experimental mutagenesis findings, compared to AlphaFold2 models.

Enhancing Cancer Classification Through Ensemble Machine Learning and Gene Selection Approaches

The vast dimensionality of gene expression data and the limited number of relevant genes necessitate the adoption of gene selection techniques. Furthermore, the choice of an efficient classifier plays a pivotal role in achieving accurate results. In this study, we employ the Minimum Redundancy Maximum Relevance (mRMR) method for gene selection, coupled with ensemble classifiers and individual classifiers like K-Nearest Neighbors (KNN) and Decision Trees (DT). A comparative analysis between two ensemble classifiers and two individual classifiers is conducted, revealing the superior performance of the ensemble classifiers. Our investigation utilizes four distinct cancer gene expression datasets to showcase the efficacy of employing ensemble classifiers and gene selection methods for cancer classification. The ensemble classifier (Bagging Classifier), in conjunction with the MRMR method selecting only the top 30 genes, achieved an impressive overall accuracy of 94% across all four employed datasets.

Integration of Blockchain & Machine Learning Approach in Sustainable Material Selection

Integration of Blockchain & Machine Learning Approach in Sustainable Material Selection

Efficient traffic-based IoT device identification using a feature selection approach with Lévy flight-based sine chaotic sub-swarm binary honey badger algorithm

Internet of Things (IoT) refers to the various devices connected to the Internet, enabling them to communicate and transmit data with each other. The rapid development of the IoT also brings security and other problems in cyberspace. In this case, device identification is a crucial tool for IoT security issues, which can detect and prevent cyber-attacks. However, device identification has some challenges on IoT traffic datasets, such as large-scale and high-dimensional sparse datasets, features prone to security vulnerabilities, and identification of IP and non-IP devices, which affects the performance of classifiers. Feature selection can be seen as an effective data preprocessing technique in IoT device identification, which may improve the performance of classification and reduce the computational complexity of IoT device identification. In this paper, we propose a traffic-based IoT device identification model using a novel wrapper feature selection approach based on an improved and efficient method, which we call Lévy flight-based sine chaotic sub-swarm binary honey badger algorithm (LS2-BHBA). Specifically, four improved factors are employed in LS2-BHBA to expand the search scope, balance the exploration and exploitation phases, and enhance the search capability. In addition, a binary mechanism is implemented to enhance the suitability of the proposed LS2-BHBA for feature selection in IoT device identification. The experimental results on several real IoT traffic datasets denote that LS2-BHBA can reduce the number of features to 10%, and the classification accuracy can reach 98%, which outperforms some classical and latest comparison algorithms in the feature selection of IoT device identification.

Coastal vulnerability assessment in urban planning: a comparative tool-based selection approach for the Egyptian context

Coastal vulnerability assessment in urban planning: a comparative tool-based selection approach for the Egyptian context

Reviewing various feature selection techniques in machine learning‐based botnet detection

SummaryMachine learning approaches are widely used for the detection and classification of emerging botnet variations due to their ability to yield more precise results compared to traditional methods. The relevancy of the features plays a major role in these detection algorithms' effectiveness. As such, the most distinctive characteristics must be extracted from a high‐dimensional dataset that is used to classify botnets. Nevertheless, we discovered that the majority of earlier studies lacked proper analysis and paid little attention to the various feature selection techniques. The main goal of this work is to investigate and assess the advantages and disadvantages of the different feature selection techniques used for botnet detection. Studies show that feature selection is a very efficient way to decrease the amount of storage and processing power required while simultaneously increasing classification accuracy. As a consequence, its application in many other fields has grown. The field of feature selection is recognized for its non‐deterministic polynomial‐time hardness; to mitigate this hardness, metaheuristic techniques have been applied. Metaheuristic algorithms are exceptionally good at performing a global search. In order to choose feature subsets optimally in the field of botnet detection, we additionally prioritize the use of metaheuristic methods. This study offers a more thorough insight of the feature selection strategies that are primarily employed by machine learning‐based botnet detection models. It also offers insights into how better feature selection approaches might be applied to strengthen botnet detection mechanisms. Additionally, it will help in understanding the limitations of existing approaches and identifying areas for improvement.

An automated speech recognition and feature selection approach based on improved Northern Goshawk optimization

<p><span lang="EN-US">Automatic speech recognition (ASR) approach is dependent on optimal speech feature extraction, which attempts to get a parametric depiction of an input speech signal. Feature extraction (FE) strategy combined with a feature selection (FS) approach should capture the most important features of the signal while discarding the rest. FS is a crucial process that can affect the pattern classification and recognition system's performance. In this research, we introduce a hybrid supervised learning using metaheuristic technique for optimum FE and FS termed Northern Goshawk optimization (NGO) and opposition-based learning (OBL). Pre-processing, feature extraction and selection, and recognition are the three steps of the proposed technique. The pre-processing is done first to lessen the amount of noise. In the FE stage, we extract features. The OBL-NGO method is used to pick the best collection of extracted characteristics. Finally, these optimised features are utilised to train the k-nearest neighbour (KNN) classifier, and the matching text is shown as the output based on these optimised characteristics of the provided input audio signal. The system's performance is outstanding, and the suggested OBL-NGO is best suited for ASR, according to the testing data.</span></p>

Spatial multi-criteria decision making approach for wind farm site selection: A case study in Balıkesir, Turkey

This study presents a spatial multi-criteria decision making (SMCDM) approach based on geographic information system (GIS) for determining the sites that have potential for wind farm installation in Balıkesir, Turkey. Topographical, structural, climatic, and environmental criteria consisting of wind speed, land cover type, slope of terrain, and distance from substations, road networks, settlements, fault lines, electricity grid, and seaports are considered in the approach. Weights, derived from evaluation involving the use of the analytical hierarchy process (AHP) questionnaire, and scored criteria maps, are used in creation of a wind farm suitability map. The results of the study demonstrate that 2.34 % of the total area under consideration is in the most suitable class, while 9.34 % is in the next most suitable class. The study introduces the seaport criterion for the first time in onshore wind turbine site selection. The findings show that considering the seaport criterion provides a more realistic approach of feasible wind turbine site selection. This is clearly seen in the 13 % decrease in the most suitable area. When the wind farm suitability map is evaluated on a district scale, Bandırma takes first place compared to other districts. A sensitivity analysis performed with equal criteria weights indicates that the results are generally reliable.

Material Selection Using Hybrid Grey Relation Analysis Approach Based on Weighted Entropy for Ranking: The Case of Helicopter Rotor Blade

Engineering design relies highly on the selection of suitable materials. Because there are many engineering materials, selecting a suitable material for a product requires a systematic selection approach. This paper provides a hybrid strategy for choosing the best material for an engineering design to give the best performance at the lowest cost based on Ashby's performance indices. Then, it ranks the result by the grey relational approach integrated with the Weighted Entropy Method to choose the optimum material for the main rotor blade of a helicopter. Different materials used for manufacturing rotor blades, such as Aluminium alloys, titanium alloys, steel, composites, and wood, have been discussed. The performance indices chosen are stiffness, strength, and fracture toughness. The performance indices proved that the composite material has excellent structural strength, stiffness, and toughness. The result shows that CFRP is the best material for manufacturing helicopter rotors, while wood and steel were the best and cheapest when the design had to be economical.

A Binary Bee Foraging Algorithm-Based Feature Selection Approach for IoT Intrusion Detection

A Binary Bee Foraging Algorithm-Based Feature Selection Approach for IoT Intrusion Detection