publications

2024

1. IntentRec: An Advanced Recommender System Leveraging User-Item Intent

   Abhishrut Vaidya and Niladri Chatterjee

   In Intelligent Computing, 2024

   Abs

   The present paper proposes a novel recommendation technique called IntentRec that focuses on modeling a movie recommender system based solely on the pattern of user ratings for the items, called intent. Unlike existing recommendation algorithms that rely on item and user descriptions, IntentRec leverages the intent behind ratings to generate more accurate recommendations. The initial intent for each user is computed as a prior-intent probability. It is represented as a distribution of ratings over rating types (i.e., 1, 2, 3, 4, 5) over all the items. Through an optimization process using a convoluted kernel layer followed by a kernelized AutoEncoder, the model calculates post-intent probability, representing the rating distribution of users and items. IntentRec addresses the challenges of variation and sparsity in user ratings by incorporating periodic updates in the user-item intents throughout the learning process. This dynamic adaptation helps the model handle varying and sparse rating data robustly. The objective of predicting missing ratings in IntentRec is to utilize similar users and similar items calibrated with comparable similarity scores. This collaborative filtering approach aids in mitigating the cold start problem, where there is limited information about new users or items. Moreover, IntentRec has fewer parameters, making it computationally efficient. The experimental results demonstrate that IntentRec performs better than the current state-of-the-art results using the RMSE metric on the Movielens and Douban datasets without requiring additional side information.