Enhancing Consumer Behavior Analysis In Retail Using Hybrid Deep Learning Models

Abstract

The driving force behind the retail firms' focusses on improving sales, customer satisfaction, and decision-making is consumer behavior. Retail organizations currently have the issue of conventional data modeling being unable to cope with the volume and complexity of current retail data. To aid in predicting consumer behavior, the researchers recommend using a combination of the two deep learning models, which are LSTM and CNN. LSTM is employed to identify temporal correlations, while CNN helps identify spatial features. This approach enhances the predictability of Consumer behaviour, including the decision to purchase. The authors used a dataset containing Retail transaction data, customer review information, and demographic information, which were highly pre-processed and processed for missing data, encoded categorical data, and processed to be used with natural language processing to handle the text data. The performance of their proposed model surpassed that of individual models of deep learning as well as traditional methods with regard to the measures used which include accuracy (88%), precision (89%), recall (87%), F1-score (91%), AUC (95%), and MSE (2%). This is as stated by the authors' ablation study in their work. Retail businesses can leverage these findings concerning Hybrid Deep Learning models, which will assist in creating targeted marketing efforts, inventory management, and demand forecasting. However, fortunately, the use of explainable AI is recommended in future research to address problems related to model interpretability and limited data.