<![CDATA[Customer churn is a major challenge in the highly competitive Indonesian Internet Service Provider (ISP) market, where companies face significant customer turnover rates impacting profitability and sustainability. This study integrates multilayer perceptron (MLP) neural networks with marketing intelligence to predict and mitigate churn effectively. The methodology includes data preparation, exploratory data analysis (EDA), and model development. EDA plays a critical role in identifying key features for churn prediction, ensuring meaningful insights into customer behavior. The model uses the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance, improving prediction performance. The final model achieved an area under the curve (AUC) of 99%, a metric that measures how well the model distinguishes between churned and non-churned customers, and an F1 score of 97%, which balances the model’s precision (accuracy of positive predictions) and recall (identification of all true churners). These findings provide actionable insights for ISPs to tailor customer retention strategies and improve business performance.]]>
Copyrights © 2024
