Small Sample Enterprise Renewal Prediction Model Based On Machine Learning

Customer churn is an important issue faced by enterprises.In particular,the cost of acquiring new customers by SaaS companies is relatively high.It requires the multidepartment cooperation of market,sales and customer service departments,and the cycle is long.After acquiring the customer,the customer pays regularly according to the time of use and function;if the usage period of the customer’s payment is short,it is likely that the customer can’t to make ends meet.Therefore,it is extremely important to make accurate predictions on whether customers are losing,and to take timely and timely retention measures for customers with different probability of loss.At the same time,improving customer retention has important value and significance for the survival and development of SaaS enterprises.The research in this paper is based on machine learning algorithm to study the renewal model of enterprise customers in the enterprise service field.Firstly,it introduced the research background and current situation analysis of customer churn;Based on the understanding of business,I collected the enterprise customer data of an Internet company who purchased a cloud product from May 1,2017 to May 1,2018 to carry out the churn prediction research.Secondly,I used exploratory data analysis methods to explore the relationship between different attributes and behaviors of customers and customer churn.Then,based on the evaluation indexes of accuracy,precision,recall,F1 value and confusion matrix,I compared the prediction performance of four classifier prediction models in the field of small sample customer churn,which are Logistic Regression,Support Vector Machine,XGBoost model and Neural Network model.Finally,the XGBoost model with the best performance is selected to output the importance of the feature and the probability of loss of all users in the prediction period.The prediction accuracy of XGBoost model on the test set reached 79%,the recall rate reached 89%,and the effect reached the goal of modeling.The renewal prediction model based on the analysis of SaaS enterprise customer data provides theoretical and methodological support for enterprises to solve customer churn and improve retention,and has certain reference value for enterprise customer relationship management.