Sustainable Product Choices Recommendation System Using Contextual Bidirectional Encoder Representations from Transformers and Multi-Layer Perceptron

Recent days, e-commerce platforms have booming tremendously as the people showing interest to buy the products from online based on the customer reviews and ratings. So, providing the best product recommendations to customer will increase global market and GDP growth of a country. The existing methods for recommendation of products had resulted in suggesting unnecessary products to customers. Therefore, in this research Contextual Bidirectional Encoder Representations from Transformers- Multi-Layer Perceptron (CoBERT-MLP) method is proposed for recommending sustainable product choices using a knowledge-based recommender system. The proposed method employed on Flipkart product dataset that provides details of different products. The data obtained from this dataset is pre-processed by utilising the Natural Language Toolkit (NLTK). These pre-processed data features are extracted by the employed CoBERT method that eliminates the irrelevant attributes from the data. These obtained relevant attributes are classified for categorizing the product choices of a customer. Finally, these classified choices are given to knowledge-based recommender system for providing most sustainable product choices to customers. The proposed CoBERT-MLP method achieved better results in terms of accuracy (99.14%), recall (99.38%), precision (98.56%), and F1-score (99.71%) when related with prevailing method Convolutional Neural Networks (CNN).