Stratified Advance Personalized Recommendation System Based on Deep Learning.

Deo, Arpit; Jaisinghani, Riya; Gupta, Sagar; Khan, Safdar Sardar; Soni, Adish; Gehlot, Kushal

A recommendation system is a refinement system that uses massive amounts of data to forecast and present user-preferred products. We employ web log files, including previously searched data, and browsing history, and transmit it to a SoftMax model in our recommendation model. We also use this data to perform user behaviour analysis using the k means clustering algorithm. Furthermore, we transfer users' feedback data, which is divided into explicit and implicit data, to the EIMNF model, which is a neural matrix model that aids us in forecasting users' preferences. Furthermore, we undertake cross domain analysis with the use of a CNN FT model, and all of the outputs created by the algorithm are referred to as intermediate recommended items, and they are delivered to an item pool to be reranked. Re-ranking improves accuracy and allows us to provide the best possible suggestions to our users. We employ a graph neural network to execute re-ranking, and the best things generated after reranking are provided to our end-user. We compared our model to various models and found that proposed model has 0.91 precision, 0.84 recall, 0.87 FMeasure and holds 91% accuracy.

RECOMMENDER systems; DEEP learning; K-means clustering

Ingénierie des Systèmes d'Information, 2023, Vol 28, Issue 1, p189

1633-1311

Academic Journal

10.18280/isi.280121