Customer churn prediction by using support vector machine

Prediction of customer churn is important due to intensive marketing competition. With the purpose of retaining customers, companies apply churn prediction models to determine the customers churn by analyzing their behavior and trying to put effort and money into retaining them. In this thesis, we develop and test a model to estimate the propensity of a customer to abandon the company in a near future. This study applies support vector machines (SVM), a machine learning technique used in binary classification. SVM was compared with different kernels: linear, radial basis function (RBF) and polynomial. The experiment was carried out in Python with machine-learning tools, along with a real database from Kaggle. Afterward, the predictive performance of three kernel show that SVM with polynomial and RBF have the best accuracy rate and provide an effective measurement for the bank's customer churn prediction (CCP). The results were shown in different evaluation measures.

• Auteur principal: Santacruz Nagua, Alfredo Fabian (author)
• Format: bachelorThesis
• Langue: eng
• Publié: 2021
• Sujets: Predicción de fuga de clientes; Máquinas de vectores de soporte; Kernel; Análisis de datos; Prediction churn; Support Vector Machines; Data analysis
• Accès en ligne: http://repositorio.yachaytech.edu.ec/handle/123456789/341