Prediction of Customer Propensity Based on Machine Learning

With the generalization of online consumption, the problem that many companies face is that they attract many visitors to website every day, but know that only a small percentage of people will actually buy the product, and most people may not even return. What the company wants to achieve is to market to real consumers, so we try to use machine learning to determine the most valuable prospects. In this article, we used logistic regression and random forest to solve the problem, and compare the prediction effects of the two models. Our model has a relatively good accuracy rate in predicting whether consumers will place an order. In addition, we also compare the prediction effects of these two models and find that as far as this problem is concerned, the accuracy of random forest is higher than that of logistic regression, while the recall rate of logistic regression, F 1 score, ROC, is higher than that of random forest. The study of this problem can provide solutions for the company's more precise marketing and the discovery of potential customers.