Predicting Customer Satisfaction in Brazil E-commerce: A Comparative Study of Machine Learning Techniques

Customers' propensity for brand loyalty, recurrent business, and positive word-of-mouth are directly influenced by the degree to which their needs are met. E-commerce happens to be one of the biggest online industries in the world which till date is not completely understood from a customer's standpoint. Brazil being a developing country is the 12
^th
largest market for e-commerce in the world and accounts for 30 % of e-commerce in Latin America and remains largely unexplored. In this study, we used machine learning (ML) techniques such as LSTM, XGBClassifier, Logistic Regression, Neural Network, and Random Forest to predict customer satisfaction using a retail dataset from Brazil. We contrasted the efficacy of these models using a variety of evaluation criteria, such as precision, recall, and Fl-score. The LSTM model had the highest accuracy with a score of 0.89, followed by the XGBClassifier with 0.83, Naive Bayes with 0.73, Logistic Regression with 0.74, Support Vector Classifier with 0.74, and Random Forest with 0.78. The superior performance of the LSTM model over other models suggests that it has superior predictive abilities.