A novel product recommendation system using hybrid deep learning-based opinion mining with collaborative filtering

To make informed decisions when buying consumer electronics, it is crucial to filter out most pertinent information by investigative industry trends, performing interviews with many key figures in the industry, and using publicly accessible data. In this work, product recommendations are performed based on opinion mining by utilizing a deep learning model. From the sentiWordNet and statistical features, a sentiment score is generated using HAN-HDLTex which is the hybridization of Hierarchical Deep Learning for Text Classification (HDLTex) and Hierarchical Attention Network (HAN) approach. Moreover, to determine product ratings, proposed Marine Predator’s algorithm- Neural Collaborative Filtering (MPA-NCF), which was designed by training NCF using Marine Predator’s algorithm (MPA), was used. Finally, a similarity function is used to merge the outcomes that involve product ratings from sentiment analysis and the optimized NCF. Finally, the experimentation analysis is performed using the metrics, namely accuracy, RMSE, MSE, MAPE, TPR, and TNR. The evaluation states that proposed technique attained maximum accuracy, TPR and TNR with the values of 0.949, 0.968, and 0.938, and also, proposed model achieved minimum MAPE, MSE and RMSE with values of 0.039, 0.019, and 0.140.