International Journal of Applied Sciences & Development
E-ISSN: 2945-0454
Volume 3, 2024
Comparative Analysis of Data Mining Classification Techniques for Prediction of Problematic Internet Shopping
Authors: ,
Abstract: As online shopping has surged, so do disorders on internet purchasing. This study aims to develop and compare predictive models that use data mining methods to predict problematic internet shopping. We used the Artificial Neural Network (ANN), CHAID with bagging, and C5.0 and compared them with traditional logistic regression to construct predictive models on a training cohort of 858 shoppers. Another cohort of 368 buyers was utilized to confirm the accuracy of the predictive model. The accuracy, sensitivity, specificity, and the ROC-AUC were used to assess the predictive performance. The C5.0 algorithm provided better accuracy in predicting PIS than the other models, indicating that C5.0 might be a practical auxiliary algorithm for predicting PIS. Our research findings cater to a comprehensive PIS prediction system, providing timely intervention and appropriate support to individuals with the PIS problem.
Search Articles
Pages: 82-88
DOI: 10.37394/232029.2024.3.7