Bidirectional Short-Term Memory with Osprey Optimization Algorithm for Automatic Recommendation System

Recommender systems are crucial to the expansion of e-commerce in various areas, and it helps users to find their interest in the Internet. Additionally, it aims to improve customer relationship management by providing customers with personalized suggestions for online goods and services to combat the growing problem of information saturated on the Internet. Although many recommender system strategies have been developed, the accurate suggestion remains difficult. To overcome this issue, an optimized BiLSTM approach is developed. Initially, the data are collected and converting column's text into numbers. These converted raw data are pre-processed utilizing z-score normalization, K-Nearest Neighbour (KNN)-based missing value imputation, and affinity propagation clustering to fill the absent values. Then, the dimensionality of the pre-processed data was reduced using quadratic discriminant analysis (QDA). Finally, a Bidirectional Short-Term Memory (BiLSTM) classifier is utilized for classification purposes. The hyperparameter present in the BiLSTM classifier is tuned using the osprey optimization algorithm (OOA). The valuation outcomes show that the suggested approach achieves 93.1% of accuracy, 85.9% of precision, and 95.9% of specificity rate for the data. In comparison to other current methodologies, the proposed approach performs better. Thus, the suggested method is the best choice for an automated recommendation system.