Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{206884,
        author = {Megha B N and Dr.Sreenivasa B.R and Dr. Nirmala C R},
        title = {A Softmax -Weighted Context Model for Sequential E-Commerce Personalization},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {2},
        pages = {3049-3054},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=206884},
        abstract = {Nowadays personalized recommendation has become a core requirement of modern e-commerce platforms, where user attention and purchasing intent shift continuously across browsing sessions. Conventional recommendation approaches, such as pure popularity ranking or static collaborative filtering, respond slowly to a user's most recent behaviour. Our paper presents the design, implementation, and evaluation of a Context-Aware Sequential Recommendation (CASR) system built as a full-stack e-commerce application using a React front end and a Flask/SQLite back end. The recommendation layer combines three complementary components such as: (i) a first-order Markov transition model that learns item-to-item purchase sequences, (ii) a category-level hybrid sequence model that maintains an exponentially-decayed, attention-weighted context vector over a user's recent cart, wishlist, and order activity, and (iii) a linear scoring classifier that predicts whether a product is likely to be a high-selling item, trained on catalogue attributes such as price, rating, discount, review count, brand, and sales rank. Our system is evaluated using rank-aware recommendation metrics (Accuracy/Hit-Rate, Precision, Recall, F1-score, NDCG, and MAP) and, for the binary sales-success classifier, Receiver Operating Characteristic (ROC) analysis with Area Under the Curve (AUC).. We conducted experiments on datasets- Amazon Product Review, Retailrocket, and Yoochoose -demonstrate significant improvements and also on a product sales dataset show that the linear sales classifier achieves an accuracy of 89.2% and an AUC of 0.71. For future work we can outline a research agenda toward learned embeddings and attention-based sequential architectures (e.g., SASRec, BERT4Rec).},
        keywords = {Context-aware recommendation, sequential recommendation, hybrid recommender system, e-commerce personalization, Markov chain recommendation, ROC-AUC evaluation, NDCG, MAP.},
        month = {July},
        }