Abstract:- With the telecommunications industry's fast growth, service providers are more focused on growing their consumerbase. The retention of existing consumers has become a huge issue to satisfy the demand of being competitive. According to apoll conducted in the telecommunications business, the cost of acquiring a new client is significantly more than the cost ofmaintaining an existing one. As a result, gathering information from the telecommunications sectors may assist in predictingthe association of clients and whether or not they would depart the organization. The telecom sectors must take the appropriatesteps to begin the acquisition of their related clients to maintain their market worth. As part of this research, we propose a newframework for the attrition prediction model, which we will then apply using Machine Learning techniques. There has been acomparison of the five algorithms in terms of efficiency and performance. This project introduces a new set of features forpredicting customer attrition, including state, account length, area code, international plan, total day minutes, total day calls,total day charge, total eve minutes, total eve calls, total eve charge, total night charge, total night charge, total intl minutes,total intl charge, number customer service call Then, based on the new attributes, the five prediction approaches (LogisticRegressions, Random Forest, Naive Bayes, Support Vector Machines, and K-Nearest Neighbours Algorithm) are used toestimate customer attrition. Finally, comparison experiments were carried out with AUC scores and ROC curves to evaluatethe new feature set as well as the five modeling strategies for customer attrition prediction in the final stage. For predictingcustomer attrition in the telecommunications industry, the new features with the five modeling methodologies are morepersuasive.KEYWORDS: Logistic Regressions, Random Forest, Naive Bayes, Support Vector Machines, K-Nearest Neighbours Algorithm