Predicting Customer Churn in a Telecommunications Company Using Machine Learning

In today’s world, if a company is not equipped with clear analysis and foreseeing, endless customer churn will occur. This industry is highly competitive and the amount of customers is fundamental to a telecom company, which can help them to gain enough profits they want initially. Some data visualization will be realized to build up some relations between factors and churn by using heat maps and matrices. At last, some algorithms will be introduced and calculated in some train data to measure which model is the best one to predict churns, such as logic regression, random forest, and decision tree. They will be compared on different occasions and each of them will be given a quantitative score. From this research, the random forest has the best performance in accuracy score and PCA curve (2110 customers). This information can be used to develop targeted retention strategies to reduce churn rates and improve customer satisfaction.