Analytical framework for predicting online purchasing behavior in Malaysia using a machine learning approach
In the fast-changing world of digital commerce, predicting online purchasing behavior is essential for improving e-commerce strategies. This study presents a post-pandemic machine learning-based analytical framework specifically designed for Malaysia’s ecommerce market. The framework addresses a gap...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | en |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/122927/1/122927.pdf https://ir.uitm.edu.my/id/eprint/122927/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In the fast-changing world of digital commerce, predicting online purchasing behavior is essential for improving e-commerce strategies. This study presents a post-pandemic machine learning-based analytical framework specifically designed for Malaysia’s ecommerce market. The framework addresses a gap in predictive analytics by combining computational techniques, consumer behavior theories, and demographic data to better understand and forecast purchasing trends. The framework uses machine learning methods, including classification, clustering, feature selection, and parameter tuning, to improve accuracy and reliability. It is organized into three phases: preliminary investigation, implementation and analysis, and validation. The descriptive analysis examines purchasing behavior through correlation and regression analyses, while the predictive model uses decision trees (J48, Random Tree, REPTree), rule-based algorithms (JRip, OneR, PART), and clustering (K-Means) to identify patterns and predict trends. Feature selection techniques, such as WrapperSubsetEval, were used to improve focus on key attributes, and parameter tuning further optimized performance. Among the three datasets analyzed (D1, D2, and D3), Dataset 3, which emphasizes psychological and emotional factors, achieved the highest accuracy and predictive performance. Validation included both machine learning evaluation and expert feedback. Machine learning results showed strong performance, with Random Tree and PART delivering the most reliable predictions. Adjustments, such as tuning K-Values and binary splits, further enhanced results. Expert feedback validated the framework’s alignment with the Howard-Sheth consumer behavior model and suggested improvements, such as including additional behavior related factors and refining segment categories. Key findings highlight the growing participation of female consumers in online shopping, signaling changing trends in Malaysia’s e-commerce landscape. This study also introduces a unique dataset on Malaysian online consumer behavior, filling a critical gap and providing valuable insights for businesses. The framework aligns with Malaysia’s Science, Technology, Innovation, and Economic (MySTIE) framework and supports SDG 8 and KEGA 2 (Digital Economy) goals. By integrating classification, clustering, and consumer behavior models, this framework offers practical tools for understanding and predicting consumer behavior. Its validation, including evaluation and expert feedback, ensures reliability and adaptability, making it a valuable resource for improving e-commerce strategies in Malaysia. |
|---|