Potential Customer Prediction of Telecom Marketing based on Machine Learning

Abstract

Telemarketing has an important application in commercial promotion, and blind product recommendation has a high failure rate. However, product recommendation to potential users can effectively reduce marketing costs and increase revenue. In this paper, 41,188 data on telemarketing from a Portuguese banking institution are selected with the classification objective of predicting whether a customer will subscribe to a time deposit account or not. The paper first preprocesses the data to fill in missing data. Secondly, this paper describes the four models used in this paper: Logistic Regression, K-Nearest Neighbor, Decision Tree and Random Forest Classifier. As well as the five-assessment metrics used to evaluate these models: accuracy, AUC value, KS value, model lift and profit. In the experimental stage, this paper uses the above four models to predict the effectiveness of bank telemarketing. And the five evaluation indexes are combined to judge the prediction effect of the models. The results show that Decision Tree and Random Forest Classifier have better prediction effect.