Paper Details
Hybrid Recommendation Model using Markov Chains and Advanced Feature Integration
Authors
Anurag Rajput, Sonu Airen, K. K. Sharma
Abstract
The need for personalized product recommendations in online grocery shopping has been further emphasized by the evolution of e-commerce platforms in recent years. Traditional recommendation paradigms, including collaborative filtering and content-based filtering, encounter various challenges, such as, data sparsity, cold-start problems, and sequential dependencies, which are crucial to consider while modeling repetitive and temporal shopping activities. To tackle these limitations, in this study, we propose a hybrid recommendation model dedicated to next-basket prediction tasks.
The approach employs Markov Chains to capture sequential and historical dependencies while leveraging features like TF-IDF; product popularity, recency, and order-specific preferences are additional influences on prediction accuracy. The hybrid modelling incorporates probabilistic techniques with contextual information to provide more personalized and relevant recommendations.
We used Next-Basket Relevance Score, next-basket coverage, and Harmonic Basket Prediction Score as metrics to evaluate the performance. Our findings show that the Markov Chain hybrid model with additional features outperforms the baseline models. To be more specific Markov + Popularity + Recency got 0.165 Next-Basket Relevance Score, 1 next-basket coverage and 0.2833 Harmonic Basket Prediction Score, which indicates a model which can deal with sparse and sequential data effectively over Markov only (Next-Basket Relevance Score: 0.095 Harmonic Basket Prediction Score: 0.1735)
This work underlines the power of being able to combine sequential modelling with context features in order to build solid recommender systems. Through this, the findings have practical implications for e-commerce platforms looking to improve customer retention and satisfaction with accurate and scalable recommendations.
Keywords
Machine Learning, Markov Chains, Next-Basket Prediction, Recommendation Systems, TF-IDF Weighting.
Citation
Hybrid Recommendation Model using Markov Chains and Advanced Feature Integration. Anurag Rajput, Sonu Airen, K. K. Sharma. 2025. IJIRCT, Volume 11, Issue 2. Pages 1-26. https://www.ijirct.org/viewPaper.php?paperId=2504048