Decision Analytics Journal最新文献

{"title":"An Improved Deep Neuro-Fuzzy and Bi-Directional Gated Recurrent Unit Model for Distributed Denial of Service Attack Detection and Mitigation","authors":"Pallavi H. Chitte ,&nbsp;Sangita S. Chaudhari","doi":"10.1016/j.dajour.2025.100608","DOIUrl":"10.1016/j.dajour.2025.100608","url":null,"abstract":"<div><div>An intrusion detection system (IDS) is integral to a robust cybersecurity infrastructure. This study presents a comprehensive and advanced methodology for monitoring and detecting unwanted or malicious activities in network-oriented environments. The proposed IDS consists of three crucial stages: pre-processing, feature extraction and detection. A refined data normalization process ensures consistent and analyzable data format in the pre-processing stage. Feature extraction involves extracting various features, including statistical features, mutual information features, information gain and improved correlation. These features train the detection model to recognize patterns associated with malicious activity. A robust hybrid classifier for the detection phase is proposed, combining the Improved Deep Neuro-Fuzzy (IDNF) and Bi-Directional Gated Recurrent Unit (Bi-GRU) models. A novel hybrid optimization algorithm called the Archimedes Updated Poor and Rich algorithm (AUPRO) is introduced to optimize this model. By blending concepts from Archimedes and Poor Rich algorithms, AUPRO achieves an optimal weight configuration, resulting in superior detection accuracy and reduced false positives. The proposed system incorporates an enhanced mitigation strategy that utilizes information gathered during the detection phase. The system initiates a BAIT mitigation process to prevent or minimize damage caused by attacks effectively following the detection process. A comprehensive comparison is conducted against state-of-the-art models to evaluate the performance of the proposed system. Metrics such as accuracy, sensitivity, specificity, false negative rate, false positive rate, precision and other relevant factors are considered in the performance study. The results demonstrate the superiority of the proposed system, showcasing its ability to provide a heightened level of security and accuracy in detecting and mitigating network attacks. Organizations can bolster their cybersecurity measures by implementing this advanced approach to intrusion detection systems and proactively safeguard their networks from potential threats and attacks.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100608"},"PeriodicalIF":0.0,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144655710","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An analytical study of design market signals through online trends and stock price dynamics","authors":"Theodoros Daglis","doi":"10.1016/j.dajour.2025.100606","DOIUrl":"10.1016/j.dajour.2025.100606","url":null,"abstract":"<div><div>This study examines the nexus between Google Trends’ collective interest in specific keywords related to technological advancements utilized in design and the stock performance of major companies in design-related sectors. Specifically, the paper examines causality patterns between Google Trends keywords and stock prices of design companies, also employing multi-fractal analysis, to test for long-term relationships. According to the results, varying impacts across keywords on stock prices are identified, with non-fungible token (NFT) exhibiting the greatest influence, followed by three-dimensional (3D) printing and computer-aided design, virtual reality (VR) displays a noteworthy impact, while artificial intelligence (AI) design and generative design indicate the least impact. The results also reveal persistent long-term relationships between the examined variables, with rich multifractal behavior indicating complex relationships, mostly balanced. The findings are important for policymakers and managers, necessitating close monitoring, especially of NFT, and for design companies to align strategies for market movements.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100606"},"PeriodicalIF":0.0,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144634329","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An analytical approach to credit risk assessment using machine learning models","authors":"Marcos R. Machado,&nbsp;Daniel Tianfu Chen,&nbsp;Joerg R. Osterrieder","doi":"10.1016/j.dajour.2025.100605","DOIUrl":"10.1016/j.dajour.2025.100605","url":null,"abstract":"<div><div>This study presents a novel Early Warning System for monitoring the credit risk of commercial customers at a large international bank headquartered in the Netherlands. Traditional early warning methods often rely on backward-looking indicators such as probability of default or loss given default, which can limit predictive performance. To address this, we investigate the effectiveness of a Watchlist-based trigger for forecasting financial distress and adverse customer migration. We assess its precision, timeliness, and sensitivity across different client status transitions. Using a rich dataset combining internal banking records and external financial information, we implement and compare several machine learning algorithms, including Linear Discriminant Analysis, Logistic Regression, Decision Trees, Support Vector Machines, Random Forest, Gradient Boosting, Extreme Gradient Boosting, and Artificial Neural Networks. To enhance model transparency and support adoption, we employ SHapley Additive exPlanations to interpret key predictors of risk. Among all models, Random Forest achieves the highest performance, demonstrating strong F1 scores, superior trigger precision, and high sensitivity to migration. It successfully anticipates 12.7% of negative client transitions and helps prevent 67.6% of cases that would otherwise result in financial losses for the bank. This research contributes a data-driven, explainable solution for proactive credit risk management and offers actionable insights to support strategic decision-making in commercial banking.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100605"},"PeriodicalIF":0.0,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144596899","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A predictive analytics framework for early detection of production halts and quality issues","authors":"Arthur Matta ,&nbsp;Luís Miguel Matos ,&nbsp;Jorge Miguel Silva ,&nbsp;Miguel Bastos Gomes ,&nbsp;André Pilastri ,&nbsp;Paulo Cortez","doi":"10.1016/j.dajour.2025.100607","DOIUrl":"10.1016/j.dajour.2025.100607","url":null,"abstract":"<div><div>This study presents a Machine Learning (ML) framework for an Ahead-of-Time (AoT) prediction of production halts and defects in particleboard manufacturing that uses only pre-production input variables. The proposed approach incorporates both Single-Task Learning (STL) and Multi-Task Learning (MTL) paradigms, which are evaluated across three production lines under two modeling strategies: Line-Specific Modeling (LSM) and Line-Agnostic Modeling (LAM). The experimental evaluation benchmarks a lightweight Logistic Regression (LogR) model against three Automated Machine Learning (AutoML) techniques: H2O AutoML, Ludwig, and a Bayesian-optimized Deep Feedforward Network (DFFN). Results show that the STL-LSM combination using LogR achieves the highest overall predictive performance. To enhance model interpretability, we apply two model-agnostic eXplainable Artificial Intelligence (XAI) techniques: SHapley Additive exPlanations (SHAP) and One-Dimensional Sensitivity Analysis (1DSA). These methods generate feature importance rankings across targets and production lines, which are evaluated using quantitative (normalized distance metrics) and qualitative measures (alignment with domain expert insights). The XAI findings reveal a strong consistency between SHAP and 1DSA, with 1DSA requiring a substantially lower computational cost. Moreover, the convergence between model-derived interpretations and expert feedback highlights the practical relevance of the proposed ML framework in supporting data-driven decision-making for particleboard production planning.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100607"},"PeriodicalIF":0.0,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144663359","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A stochastic programming framework for pricing and market share optimization in retail systems","authors":"Muhammed Sütçü ,&nbsp;Barış Yıldız","doi":"10.1016/j.dajour.2025.100604","DOIUrl":"10.1016/j.dajour.2025.100604","url":null,"abstract":"<div><div>This study examines a scenario where a manufacturer owns one retailer and collaborates with an independent retailer to sell a single product with Poisson demand over a multi-period selling horizon. The manufacturer protects the independent retailer’s profitability through price protection and mid-life and end-of-life return opportunities. The retailers are allowed to place replenishment orders throughout the selling horizon. The manufacturer-controlled and independent retailers manage their stocks through order-up-to policies and hybrid policies comprising order-up-to and dispose-down-to policies, respectively. We employ stochastic programming techniques to construct models to determine the manufacturer’s optimal pricing strategy. Retail Fixed Markdown (RFM) policy is assumed to determine the retail price at which the independent retailer sells the product. We also consider the impact of retail prices on the retailers’ market shares, which influence the mean demand observed by each retailer. We propose a modified version of the Stochastic Dual Dynamic Programming (SDDP) algorithm to determine the manufacturer’s approximately optimal pricing strategy. Then, we examine how price protection contract parameters affect the manufacturer’s approximately optimal pricing strategy and the retailers’ expected total profits. We also make comments on the selection of ideal values for the parameters.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100604"},"PeriodicalIF":0.0,"publicationDate":"2025-07-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144631331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A machine learning approach to consumer behavior in supermarket analytics","authors":"Tasos Stylianou ,&nbsp;Aikaterina Pantelidou","doi":"10.1016/j.dajour.2025.100600","DOIUrl":"10.1016/j.dajour.2025.100600","url":null,"abstract":"<div><div>The rapid advancement of Big Data technologies has significantly influenced multiple sectors, with the retail industry being a key impact area. This study explores the relationship between Big Data analytics and consumer behavior, focusing on supermarket transaction data to extract macroeconomic insights. Using a dataset of over two million records from a multinational supermarket chain, the research employs a suite of advanced analytical techniques, including the Apriori algorithm for association rule mining, K-Means clustering for customer segmentation, collaborative filtering for recommendation systems, and AutoRegressive Integrated Moving Average (ARIMA) for time series forecasting. The study identifies purchasing patterns, segment-specific preferences, and temporal shopping behaviors. The results reveal distinct associations among frequently co-purchased products, clear segmentations based on shopping habits, and predictive trends in consumer demand, offering valuable input for marketing, inventory management, and policy-making. Beyond the operational insights, this study highlights the potential of transactional data to reflect broader economic shifts, such as changes in consumption patterns during periods of economic uncertainty, thus linking consumer micro-behaviors to macroeconomic indicators. The novelty of this work lies in its integration of multiple machine learning techniques within a unified framework that connects retail analytics to economic policymaking, thereby extending the application of Big Data from commercial strategy to public economics.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100600"},"PeriodicalIF":0.0,"publicationDate":"2025-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144548972","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A refined k-means clustering approach for optimizing urban police facility placement","authors":"Fernando Keller ,&nbsp;Thyago Celso Cavalcante Nepomuceno ,&nbsp;Victor Diogho Heuer de Carvalho","doi":"10.1016/j.dajour.2025.100603","DOIUrl":"10.1016/j.dajour.2025.100603","url":null,"abstract":"<div><div>This study investigates spatial crime patterns in the municipality of Matelândia, Brazil, with the aim of optimizing the placement of police facilities through an enhanced clustering approach. We use an unsupervised machine learning technique, and we introduce a refined <em>k</em>-Means methodology (called <em>k</em>-Means-<em>c²</em>), which identifies a centroid of centroids representing an optimal location derived from the central points of individual crime clusters. Contrary to the traditional clustering methods that treat all data points equally, our approach incorporates differentiated crime severity by assigning varying levels of importance to four distinct crime types: theft, robbery, domestic violence, and rape. Each crime category is analyzed independently, with tailored cluster counts selected to reflect both frequency and seriousness. A global centroid is then computed to prioritize regions with a higher concentration of severe crimes. We use post-pandemic crime data and empirically analyze whether the current military police base is optimally located. We then examine the potential need for an additional facility and assess the alignment between existing infrastructure and the spatial distribution of criminal activity. This research contributes to data-driven public safety planning by combining crime analytics with facility location optimization, offering a scalable and practical framework for urban law enforcement agencies.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100603"},"PeriodicalIF":0.0,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144570847","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A data-driven approach to customer lifetime value prediction using probability and machine learning models","authors":"Albert Wong,&nbsp;Andres Viloria Garcia,&nbsp;Yew-Wei Lim","doi":"10.1016/j.dajour.2025.100601","DOIUrl":"10.1016/j.dajour.2025.100601","url":null,"abstract":"<div><div>Customer lifetime value is an important marketing metric and has applications in market segmentation, strategy development, and direct marketing programs, especially when customers are not under contract. In this research, we demonstrate the prediction of the lifetime value of patients in a health service portfolio in two separate ways. The probability of a patient being alive and their value in the coming evaluation period are first predicted using a probability model that has been well-established in the marketing community. We then use several machine learning algorithms to perform the same task. The results of these two approaches are compared in terms of accuracy to gain insight into their respective strengths and weaknesses. We believe that the work is one of the first attempts to gain an understanding of the use of machine learning algorithms in this important marketing issue. The results showed that the probability model performs better than the machine learning models, probably due to the assumption required in the probability calculations. It is therefore recommended that an essential step in applying these software approaches is to verify the validity of the key assumption of regularity. In addition, in future studies, consideration should be given to a larger dataset with demographic variables beyond age and gender that were used in this study. Developing specific ML models for dealing with zero-inflated data, which is an inherent feature of customer lifetime data, will also be helpful.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100601"},"PeriodicalIF":0.0,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144570849","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An integrated deep learning approach for predictive vehicle maintenance","authors":"Abderrachid Errezgouny ,&nbsp;Youness Chater ,&nbsp;Carlos D. Barranco González ,&nbsp;Abdeljabbar Cherkaoui","doi":"10.1016/j.dajour.2025.100597","DOIUrl":"10.1016/j.dajour.2025.100597","url":null,"abstract":"<div><div>In the automotive sector, vehicle data gathered through On-board Diagnostics (OBD) systems offers continuous insights into vehicle health status and performance. Leveraging this data for predictive maintenance can significantly reduce unplanned failures, enhance safety, and extend vehicle lifespan. This paper proposes a novel hybrid model for Predictive Maintenance (PdM), that integrates Long Short-Term Memory (LSTM) neural networks with K-means clustering to analyze unlabeled time-series data from OBD systems. Our main contribution is to integrate an unsupervised deep learning approach that effectively captures temporal dependencies and clusters operational patterns to predict engine condition with high accuracy, addressing the common challenge of unlabeled vehicle datasets. The model achieves state-of-the-art prediction performance with a 97.5% R² score of the selected feature, demonstrating its strong generalization and reliability in different domain applications. Compared to standalone LSTM, Gated Recurrent Units (GRUs) and Recurrent Neural Networks (RNNs) models, our hybrid approach outperforms traditional methods across all tested metrics, marking a significant advancement in predictive maintenance for vehicular systems. This work paves the way for smarter, real-time diagnostics in next-generation vehicles.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100597"},"PeriodicalIF":0.0,"publicationDate":"2025-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144518799","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An integrated analytics model for supplier selection and order allocation with machine learning and multi-criteria optimization","authors":"Enty Nur Hayati,&nbsp;Wakhid Ahmad Jauhari,&nbsp;Retno Wulan Damayanti,&nbsp;Cucuk Nur Rosyidi","doi":"10.1016/j.dajour.2025.100599","DOIUrl":"10.1016/j.dajour.2025.100599","url":null,"abstract":"<div><div>Sustainable Supplier Selection and Order Allocation (SSSOA) are critical strategic decisions in supply chain management. The decision-making process becomes complex under uncertainty, especially in a multi-supplier, multi-item, and multi-period environment. This study proposes a four-stage framework to address the SSSOA planning problem. In the first stage, machine learning techniques with the Autoregressive Integrated Moving Average (ARIMA) method are used to determine future product demand. In the second stage, Life Cycle Analysis (LCA) is used to determine the environmental impact of purchased drugs. In the third stage, a fuzzy supplier evaluation model based on the Best-Worst Method (BWM)-Additive Ratio Assessment (ARAS) method is used to determine supplier scores. Finally, a fuzzy probabilistic multi-objective mixed integer linear programming model is developed to determine the optimal drug order. This model aims to minimize the total purchase cost, probabilistic defects, and environmental impacts and maximize the total purchase value of the order allocation. The weighted sum method is used to solve the model. The application of the proposed framework is tested using a real dataset from a teaching hospital in Surakarta, Indonesia. The results show that this model can minimize the purchasing cost by 168.11 million Indonesian Rupiah (IDR), optimizing the total allocation value by 6,519.731 units. Sensitivity analysis of parameters such as holding cost, <span><math><mi>α</mi></math></span> value, and supplier capacity reveals that significant changes in these parameters substantially affect the total purchasing cost and order allocation. The implications of this study include improving planning accuracy, reducing environmental impacts, and optimizing supplier selection amid uncertainty, with potential applications in various other industrial sectors.</div></div>","PeriodicalId":100357,"journal":{"name":"Decision Analytics Journal","volume":"16 ","pages":"Article 100599"},"PeriodicalIF":0.0,"publicationDate":"2025-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144548975","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}