Selection Method Research Articles

Using spatial analysis and variable selection of traffic violations to determine crash frequency

Numerous studies have determined a relationship between traffic violations and crash risk. However, not only have the results been inconsistent, but the effects of minor traffic violations have not been studied extensively. In addition, the characteristics of violation data (spatial autocorrelation) are widely ignored. Hence, various types of traffic violations and spatial variable selection methods (stepwise regression, adaptive LASSO, random forest, and Boruta) were employed to select key violation variables that could be used to predict crash frequency. Then, the study used Geographically Weighted Negative Binomial Regression (GWNBR) model to determine their varying associations with crash frequency. The results indicated that Boruta had the lowest error in variable selection, and GWNBR was the best-fitting model with a heterogeneous association between crash frequency and traffic violations. The study found that areas with a higher percentage of specific behaviour-related violations and other minor violations (illegal parking) were related to higher crash frequency.

A multi-stage feature selection method to improve classification of potential super-agers and cognitive decliners using structural brain MRI data-a UK biobank study.

Cognitive aging is described as the age-related decline in areas such as memory, executive function, reasoning, and processing speed. Super-Agers, adults over 80 years old, have cognitive function performance comparable to middle-aged adults. To improve cognitive reserve and potentially decrease Alzheimer's disease (AD) risk, it is essential to contrast changes in regional brain volumes between "Positive-Agers" who have superior cognitive performance compared to their age peers but are not 80 years old yet and aging adults who show cognitive decline (i.e., "Cognitive Decliners"). Using longitudinal cognitive tests over 7-9 years in UK Biobank, principal component analysis (PCA) was first applied to four cognitive domains to create a general cognition (GC) composite score. The GC score was then used to identify latent cognitive groups. Given cognitive groups as the target variable and structural magnetic resonance imaging (sMRI) data and demographics as predictors, we developed a multi-stage feature selection algorithm to identify the most important features. We then trained a Random Forest (RF) classifier on the final set of 54 selected sMRI and covariate predictors to distinguish between Positive-Agers and Cognitive Decliners. The RF model achieved an AUC of 73%. The top 6 features were age, education, brain total surface area, the area of pars orbitalis, mean intensity of the thalamus, and superior frontal gyrus surface area. Prediction of cognitive trajectory types using sMRI may improve our understanding of successful cognitive aging.

Exploring the role of health-related quality of life measures in predictive modelling for oncology: a systematic review.

Health related quality of life (HRQoL) is increasingly assessed in oncology research and routine care, which has led to the inclusion of HRQoL in prediction models. This review aims to describe the current state of oncological prediction models incorporating HRQoL.A systematic literature search for the inclusion of HRQoL in prediction models in oncology was conducted. Selection criteria were a longitudinal study design and inclusion of HRQoL data in prediction models as predictor, outcome, or both. Risk of bias was assessed using the PROBAST tool and quality of reporting was scored with an adapted TRIPOD reporting guideline.From 4747 abstracts, 98 records were included in this review. High risk of bias was found in 71% of the publications. HRQoL was mainly incorporated as predictor (78% (55% predictor only, 23% both predictor and outcome)), with physical functioning and symptom domains selected most frequently as predictor. Few models (23%) predicted HRQoL domains by other or baseline HRQoL domains. HRQoL was used as outcome in 21% of the publications, with a focus on predicting symptoms. There were no difference between AI-based (16%) and classical methods (84%) in model type selection or model performance when using HRQoL data.This review highlights the role of HRQoL as a tool in predicting disease outcomes. Prediction of and with HRQoL is still in its infancy as most of the models are not fully developed. Current models focus mostly on the physical aspects of HRQoL to predict clinical outcomes, and few utilize AI-based methods.

A graph partitioning-based hybrid feature selection method in microarray datasets

A graph partitioning-based hybrid feature selection method in microarray datasets

The Development of a Selective Colorimetric Sensor for Cu2+ and Zn2+ in Mineral Supplement with Application of a Smartphone Paper-Based Assay of Cu2+ in Water Samples

Herein, we developed a colorimetric method for the determination of Cu2+ and Zn2+ using NBD-G as a novel selective metal sensor. NBD-G was easily synthesized by a nucleophilic substitution reaction between 4-chloro-7-nitro-2,1,3-benzoxadiazole (NBD-Cl) and Girard’s Reagent P. The NBD-G solution is yellow, but when it reacts with Cu2+ and Zn2+, its color changes selectively to red (510 nm) and orange (480 nm), respectively. NBD-G was used as a sensor for Cu2+ and Zn2+, showing a high sensitivity down to 0.77 µM for Cu2+ and 1.66 µM for Zn2+. NBD-G could determine both metals simultaneously; thus, it was applied to determine them in multimineral supplements, which showed excellent recoveries. Next, a filter paper impregnated with NBD-G was prepared as a test paper, and a simple, selective, and rapid onsite method for quantifying Cu2+ was developed as, interestingly, the paper showed no change upon the addition of Zn2+. Next, Cu2+ could be quantified with high selectivity and accuracy by photographing the color change with a smartphone camera and processing the image with Image J. The detection limit for Cu2+ using this method was 3.9 µM. Finally, the NBD-G test paper method was able to satisfactorily quantify Cu2+ spiked into the rainwater.

Authentication of Coffea arabica: Comparative Multispectral Discrimination by NIR and UV–Vis Spectroscopy Aided by Artificial Neural Networks and Data Fusion

ABSTRACTCoffea arabica L. (Arabica) is considered the highest quality coffee species and provides the majority of the worldwide production. Among its groups are rare and expensive coffees; consequently, Arabica is subject to fraudulent practices. Currently, a selection of methods allow to precisely discriminate between the primary commercial coffee species (Arabica and Robusta). However, Arabica coffees offer a very restricted genetic diversity, and the authentication of its groups by spectroscopy has not yet been demonstrated to be feasible. This study aimed to step beyond the current state‐of‐the‐art, evaluating the possibility of a multispectral approach to discriminate shelf‐ready coffee beans of the most significant Arabica groups, Typica and Bourbon. Spectral analysis was performed by a benchtop near‐infrared (NIR; 1000–2500 nm), a benchtop ultraviolet–visible (UV–Vis; 200–1000 nm), and a handheld Vis‐NIR (350–2500 nm) spectrometer, hyphenated with nonlinear classification techniques, including artificial neural networks (ANNs). Additionally, low‐level (LLDF) and mid‐level (MLDF) data fusion as well as variable selection (VS) methods were evaluated in their predictive performance. This study successfully demonstrates that closely related Arabica groups are discriminable from each other using a rapid multispectral approach, including direct on‐site analysis. The NIR region provided precise classification of the Arabica coffees (94%–98% accuracy), the Vis‐NIR (88%–93 %), and the UV–Vis (82%–93 %) region showed also good discrimination but left room for improvement. However, an LLDF of the Vis‐NIR and the UV–Vis region in combination with VS proved to be a potent tool to further refine the authentication of coffee groups and showed very good classification accuracies (91%–97%).

Skin Cancer Detection using CNN (Convolution Neural Network) with AI Medical Assistant

Skin cancer is a very common disease, and with numerous morphological forms and characteristics that are similar between malignant and benign skin lesions, makes the diagnosis pretty difficult. Recent trends in deep learning, with CNNs being the first to enter this domain, have highly facilitated the classification and detection of skin cancer, even in its nascent stages, from images. A plethora of such research works proposed novel architectures for CNNs that may possibly improve the accuracy in the diagnosis of the lesion given as against those belonging to seborrheic keratosis, nevus, or melanoma. The best model, according to the works on these authors, was the multi-layer deep learning model known as SkinLesNet, which performed better than established models such as ResNet50 and VGG16 in all datasets. More extensive studies present a DCNN model that addresses the issue of class imbalance and is stronger than transfer learning models, such as DenseNet and MobileNet. This model achieved strong accuracy on several instances: 98.5% with HAM10000 and 97.1% with ISIC-2019 data. Optimizations in model selection methods: In addition, there existed optimizations in model selection. This was through the application of the multiple-criteria decision making approaches. For instance, the method RAPS can be taken as an example. Through the application of this method, a very significant increase in accuracy was achieved. For optimizing the use of feature selection using Inception V3 in determining diagnosis for oral cavity squamous cell carcinoma, BPSO showed an accuracy of 96.3%. Such breakthroughs underpin therefore the capability of deep learning and AI techniques in raising diagnostic accuracy while cutting costs and making a chance for early intervention of skin and other cancers.

Multi-objective particle swarm optimization based on particle contribution and mutual information for feature selection method

Multi-objective particle swarm optimization based on particle contribution and mutual information for feature selection method

Attribute selection methods based on graph theory in updated formal contexts

Attribute selection methods based on graph theory in updated formal contexts

Physics-Assisted Machine Learning Models for Predicting Pool Boiling Critical Heat Flux for Cryogenic Fluids

Abstract Cryogenic fluid management is highly crucial for NASA?s space missions and one challenge identified by NASA is the tank fill process for pure cryogenic fluids. To date, no predictive tool can accurately estimate the pool boiling curve during this process. We aim to address this gap by first accurately predicting critical heat flux. We use both traditional machine learning and physics-assisted machine learning methods to predict critical heat flux on a large dataset of 1322 datapoints from 53 published studies with a wide range of operational conditions and surface properties. While the traditional machine learning approach used no prior knowledge in modeling, the physics-assisted machine learning approach assumes the correlation form of the Zuber instability model [1,2] and predicts the multiplier for the hydrodynamic instability term to estimate the critical heat flux. Two groups of input features are compared including one with fluid-only properties and another with both fluid and solid surface properties. Various feature selection methods are employed to reduce the redundant features, and multiple machine learning models are used for modeling the dataset. Physics-assisted machine learning methods predict much better than traditional machine learning methods. The inclusion of solid properties further improves the predictions of critical heat flux. The best predictions are obtained using the extreme gradient boosting model based on the physics-assisted machine learning framework resulting in a mean absolute percentage error of 8.88% which is far better than the semi-empirical correlations.

Research on Low-Carbon Site Selection Decision Model and Algorithm for Green Logistics

This work proposes a low-carbon site selection decision model and algorithm for green logistics. The focus is to minimize carbon emissions while optimizing the distribution network. With the development of smart shared logistics and increasing environmental awareness, there is an urgent need to tackle issues in urban logistics, including pollution, resource inefficiency, high carbon energy usage, and greenhouse gas emissions. By integrating geographic, environmental, and operational data, the proposed site selection model incorporates multi-objective optimization to balance carbon efficiency with cost-effectiveness in logistics site selection. We employ a hybrid algorithm combining Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) to solve the complex decision-making process. The model is tested using real-world logistics data, and the results demonstrate significant reductions in carbon emissions and transportation costs when compared to conventional logistics site selection methods. Further sensitivity analysis reveals that the model is robust under varying transportation demands and geographic constraints. Also, we highlight the role of green logistics in achieving corporate sustainability goals aligning with global carbon neutrality initiatives. We believe this model offers a practical approach to fostering environmentally responsible supply chains and will contribute to the advancement of sustainable logistics operations, helping organizations make informed, eco-friendly decisions in site selection.

Industry 4.0 technology implementation in manufacturing: a selection method and real case applications

The pressing necessity for digitalisation in industrial plants, driven by Industry 4.0 national initiatives and heightened global competition, underscores the urgency for companies to initiate digital transformation projects. Despite this urgency, the academic literature lacks comprehensive guidance on models specifically dedicated to the selection of digital technologies. This article addresses this gap by proposing a multi-criteria decision-making model, grounded in a methodological framework, for the systematic selection of digital technologies in the manufacturing sector. The proposed model combines fuzzy logic and the analytic hierarchy process (AHP) and incorporates a well-established classification of digital technologies. The model is able to select the single best candidate technology as well as the best candidate group of technologies that share the same purpose. In this way, the model tries to capture the interconnection element that is at the core of the digitalisation concept. To test its validity, the model was applied in two manufacturing companies operating in distinct production sectors. One of these companies was undergoing a digitalisation process in its plants, providing an additional basis for comparing the results of the proposed model.

Is It Worth Imitating Other Investors’ Determinants of Effective Investment Strategies Among Students

Purpose: This study examines the investment strategies of Generation Z, focusing on how they define and implement these strategies. It explores the factors influencing their decisionmaking, such as risk aversion and time horizon, and investigates the impact of social interactions on their investment choices. Design/ methodology/approach: The research involved a total of 626 participants across two academic years (218 in 2022 and 408 in 2023), who developed and executed virtual investment strategies over a seven-week period using GPWtrader, an educational platform by the Warsaw Stock Exchange Foundation. Weekly meetings were held where participants presented their strategies, market observations, and results. After the investment period, a quantitative study was conducted with 314 students from the total sample (106 in 2021/22 and 208 in 2022/23), collecting data on their investment goals, instrument selection, and strategy adaptation methods. Findings: The main finding is that any strategy turned out to bring a better return rate than a strategy that was constantly shifting. Also, different strategies were adopted and many students were imitating the strategies of their peers, suggesting that the social aspect of the investment decision-making process should not be underestimated. Research limitations/implications: The study’s main limitations are its focus on a single, homogeneous group and its sample size, which limited the use of advanced statistical analyses. The authors plan to extend the research to compare the students’ returns with market portfolio returns, analyze correlations between strategies and outcomes, and include other sub-populations of Generation Z. Originality/value: This paper extends the perspective on the investment behaviors of Generation Z, emphasizing the importance of consistency in strategy execution and the influence of social interactions. It makes a significant contribution to understanding generational differences in financial decision-making, offering practical implications for financial education and investment training.

Genotype-Driven Phenotype Prediction in Onion Breeding: Machine Learning Models for Enhanced Bulb Weight Selection

Onions (Allium cepa L.) are a globally significant horticultural crop, ranking second only to tomatoes in terms of cultivation and consumption. However, due to the crop’s complex genome structure, lengthy growth cycle, self-incompatibility, and susceptibility to disease, onion breeding is challenging. To address these issues, we implemented digital breeding techniques utilizing genomic data from 98 elite onion lines. We identified 51,499 high-quality variants and employed these data to construct a genomic estimated breeding value (GEBV) model and apply machine learning methods for bulb weight prediction. Validation with 260 new individuals revealed that the machine learning model achieved an accuracy of 83.2% and required only thirty-nine SNPs. Subsequent in silico crossbreeding simulations indicated that offspring from the top 5% of elite lines exhibited the highest bulb weights, aligning with traditional phenotypic selection methods. This approach demonstrates that early-stage selection based on genotypic information followed by crossbreeding can achieve economically viable breeding results. This methodology is not restricted to bulb weight and can be applied to various horticultural traits, significantly improving the efficiency of onion breeding through advanced digital technologies. The integration of genomic data, machine learning, and computer simulations provides a powerful framework for data-driven breeding strategies, accelerating the development of superior onion varieties to meet global demand.

The Evolution and Role of Molecular Tools in Measuring Diversity and Genomic Selection in Livestock Populations (Traditional and Up-to-Date Insights): A Comprehensive Exploration

Distinctive molecular approaches and tools, particularly high-throughput SNP genotyping, have been applied to determine and discover SNPs, potential genes of interest, indicators of evolutionary selection, genetic abnormalities, molecular indicators, and loci associated with quantitative traits (QTLs) in various livestock species. These methods have also been used to obtain whole-genome sequencing (WGS) data, enabling the implementation of genomic selection. Genomic selection allows for selection decisions based on genomic-estimated breeding values (GEBV). The estimation of GEBV relies on the calculation of SNP effects using prediction equations derived from a subset of individuals in the reference population who possess both SNP genotypes and phenotypes for target traits. Compared to traditional methods, modern genomic selection methods offer advantages for sex-limited traits, low heritability traits, late-measured traits, and the potential to increase genetic gain by reducing generation intervals. The current availability of high-density genotyping and next-generation sequencing data allow for genome-wide scans for selection. This investigation provides an overview of the essential role of advanced molecular tools in studying genetic diversity and implementing genomic selection. It also highlights the significance of adaptive selection in light of new high-throughput genomic technologies and the establishment of selective comparisons between different genomes. Moreover, this investigation presents candidate genes and QTLs associated with various traits in different livestock species, such as body conformation, meat production and quality, carcass characteristics and composition, milk yield and composition, fertility, fiber production and characteristics, and disease resistance.

AlzGenPred - CatBoost-based gene classifier for predicting Alzheimer’s disease using high-throughput sequencing data

AD is a progressive neurodegenerative disorder characterized by memory loss. Due to the advancement in next-generation sequencing, an enormous amount of AD-associated genomics data is available. However, the information about the involvement of these genes in AD association is still a research topic. Therefore, AlzGenPred is developed to identify the AD-associated genes using machine-learning. A total of 13,504 features derived from eight sequence-encoding schemes were generated and evaluated using 16 machine learning algorithms. Network-based features significantly outperformed sequence-based features, effectively distinguishing AD-associated genes. In contrast, sequence-based features failed to classify accurately. To improve performance, we generated 24 fused features (6020 D) from sequence-based encodings, increasing accuracy by 5–7% using a two-step lightGBM-based recursive feature selection method. However, accuracy remained below 70% even after hyperparameter tuning. Therefore, network-based features were used to generate the CatBoost-based ML method AlzGenPred with 96.55% accuracy and 98.99% AUROC. The developed method is tested on the AlzGene dataset where it showed 96.43% accuracy. Then the model was validated using the transcriptomics dataset. AlzGenPred provides a reliable and user-friendly tool for identifying potential AD biomarkers, accelerating biomarker discovery, and advancing our understanding of AD. It is available at https://www.bioinfoindia.org/alzgenpred/ and https://github.com/shuklarohit815/AlzGenPred.

Effective RUL prediction of the milling tool based on Shannon entropy feature selection method and squared exponential-Gaussian process regression model

Effective RUL prediction of the milling tool based on Shannon entropy feature selection method and squared exponential-Gaussian process regression model

A swelling ‎pressure prediction model based on boruta feature selection mila region

Swelling soils are problematic soil types that are prevalent across the globe. It ‎was noted that the costs ‎associated with damages caused by distended soils are ‎relatively high and this issue cannot be ignored. ‎Swelling pressure is a ‎fundamental parameter in the prediction of the swelling capacity of expansive ‎soils. In ‎machine learning, feature selection methods allow us to reduce computation time, enhance prediction accuracy, ‎and gain a deeper comprehension of the ‎data. In this paper, the Boruta algorithm is used to remove iteratively ‎the features ‎which are proved by a statistical test to be less relevant from 15 geotechnical ‎variables to predict ‎swelling pressure. The remaining variables are ‎inputs of a neural networks model (ANN). Results based on R ‎squared ‎determination coefficient, RMSE, MAPE, MSE, and RRSE show an ‎improvement of the neural model ‎by considering selected features by the Boruta ‎algorithm compared to the one without feature selection.‎ This approach highlights the effectiveness of feature selection in enhancing machine learning models for geotechnical applications.

An improving secure communication using multipath malicious avoidance routing protocol for underwater sensor network

The Underwater Sensor Network (UWSN) comprises sensor nodes with sensing, data processing, and communication capabilities. Due to the limitation of underwater radio wave propagation, nodes rely on acoustic signals to communicate. The data gathered by these nodes is transmitted to coordinating nodes or ground stations for additional processing and analysis. The characteristics of UWSN with underwater channels make them vulnerable to malicious attacks. UWSN communication networks are particularly susceptible to malicious attacks owing to high bit error rates, significant propagation delay variations, and low bandwidth. Moreover, because of the challenging and erratic underwater conditions, limited bandwidth, slow data transmission speed, and power constraints of underwater sensor nodes establishing secure communication in UWSN presents a significant challenge. To address the issues mentioned above, we have introduced the Multipath Malicious Avoidance Routing Protocol (M2ARP) and Foldable Matrix based Padding Rail Fence Encryption Scheme (FM-PRFES) methods to enhance secure communication in UWSNs. The proposed FM-PRFES method encrypts the input data to prevent unauthorized access during transmission within the network. Subsequently, the proposed Energy Efficiency Node Selection (EENS) method is used to identify the significant nodes in the network. Additionally, the Cuckoo Search Optimization (CSO) method is utilized to select the Cluster Head (CH) for data transmission. Subsequently, M2ARP is employed to analyze various routes and avoid adversarial nodes in the network. As a result, the proposed experimental analysis yields more efficient results regarding security, Packet Delivery Ratio (PDR), and throughput performance than traditional approaches.

Comparison of different machine learning methods in river streamflow estimation using isovel contours and hydraulic variables

ABSTRACT This study evaluates several machine learning (ML) models for estimating streamflow in the Osage River, Missouri, USA, and the Severn River, UK, using hydraulic input variables. These input variables are the flow section area (A), wetted perimeter (Pw ), water surface width (W), and velocity parameter (U) derived from isovel contours. Before using these models, the influential variables are selected using the sequential feature selection (SFS) method to reduce model complexity while maintaining performance. The two most important input variables identified were A and U, reducing the number of inputs from four to two. The results showed that the river’s flow conditions affect the accuracy of ML models used for estimating streamflow. Linear models such as MLR and ANFIS perform better in steady flow conditions (Osage River). In contrast, decision tree-based and non-linear models such as SVR with a radial basis kernel function (SVR_RBF) are better for unsteady flow conditions (Severn River). The finding suggests simpler models outperform complex deep-learning approaches (such as LSTM) for estimating streamflow by hydraulic variables. By selecting appropriate and efficient ML models, hydrometer stations can significantly improve the accuracy of streamflow estimation, leading to more informed decision-making in water management.