إستخدام خوارزميات التعلم الآلي للتنبؤ بتسرب العملاء لشركات الإتصالات في اليمن

Abstract

Keeping customers satisfied is truly essential for saying that business is successful especially in the telecom. Many companies experience different techniques that can predict churn rates and help in designing effective plans for customer retention since the cost of acquiring new customers is much higher than the cost of retaining the existing customers. In this thesis, three machine learning algorithms have been used to predict churn namely Support Vector Machine, Naïve Bayes and Decision Trees using two benchmark datasets (IBM Watson and cell2cell) and one real dataset from Yemen Telecom Company. The models performance has been measured by Area Under Curve(AUC) and they scored 0.82, 0.87, 0.77 respectively for IBM dataset and 0.98, 0.99, 0.98 respectively for cell2cell dataset. The proposed models obtained higher accuracy than the previous studies that used the same benchmark datasets. Then, dataset from Yemen provider have been preprocessed and tested for 3 years of 2012 to 2015’s year. The same proposed models scored 0.94, 0.97, 0.99 respectively on this dataset. After that, the models trained on another dataset for 2016 to 2018 and they scored accuracy close to the previous years. Then, the churn has been predicted for unseen dataset for the year 2019. The research then recommended some points for retaining telecom customers according to the analyzed data and the features importance based on Decision Tree model. Using the models, the telecom companies can predict and prevent the churn, then retain their customers by analyzing the customer’s records continuously, and provide them offers based on their preferred services only.