A DEEP LEARNING FRAMEWORK TO IMPROVE CUSTOMER RETENTION

Abstract

Customer retention is the most significant challenge going through the companies. Acquiring a new customer is a costly process, so organizations should keep their existing loyal customers to increase their profit and revenue and avoid customer churn. The deep learning techniques significantly impact improving and predicting customer retention. There are a lot of scientific papers that use traditional machine learning techniques to improve customer retention, but these techniques face many challenges in terms of accuracy. For this purpose, the research community began to use deep learning techniques to improve customer retention, and these techniques increased the accuracy. However, they did not focus on improving dataset quality. Therefore, this research aims to introduce a framework to improve customer retention in telecommunications companies by using deep learning techniques. In addition, this research focused on improving dataset quality using data preprocessing techniques, such as noise removal, fill null values, feature scaling (normalization, standardization), discretization, and dimensionality reduction to maximize the accuracy and get better results. The experiment is performed on a telecom dataset obtained from Kaggle called Cell2Cell. Our research achieved more accurate results, especially in predicting the loss of customers and improving customer retention by applying data quality techniques with Deep Neural Network (DNN). Our proposed model achieved good performance, demonstrating achieved 99.80% accuracy. The early identification of customers who are leaving the company and going to another vendor can assist the administration in offering them fantastic deals and lower pricing.