Online Food Delivery Customer Churn Prediction: A Quantitative Analysis on the Performance of Machine Learning Classifiers

Securing current customers is extremely necessary than earning new customers in a market that is expanding. To trace customer churn, a reliable churn prediction paradigm is required. Customer churn is the process through which people switch from one firm to another or break off contact with the company. This decision is driven by a variety of influences. It is critical for companies to acknowledge each one so that they can encourage customers to stay over. This is accomplished by regularly conducting surveys regarding customer satisfaction and analyzing the responses. Applying appropriate modelling approaches is a vital component of predicting customer churn. Predominantly, this study evaluates several machine learning models and also an incorporated model that aids in predicting customer churn where the data collected from Bengaluru regions in India about online food delivery is prioritized. In order to make better predictions using machine learning, a variety of general classifiers and ensemble classifiers are used and their degree of functionality are assessed by determining their accuracy and area under the ROC curve. According to the AUC scores obtained for the individual classifiers, the Naïve Bayes and random forest classifiers rank first with the same AUC score of 0.952. After dealing with this case, the results show that the random forest classifier outperforms all other models used.