Machine Learning Approach for Complaint Prediction in the Telecom Industry

The quality of service of the individual user of telecommunication services is one of the most critical criteria in the telecom industry that affects customer satisfaction. This paper aims to predict customer complaints before escalation and validate current escalated complaints in fixed broadband from telecommunication companies via historical and real-time customer data, including service logs and previous complaints, to optimize resource allocation and maximize customer satisfaction as the challenges are related to the volume of physical complains and churn. The dataset is real-world data provided by a single Egyptian mobile network operator and consists of 12 features pertaining to 100,000 customers. We represent 12 predictive models developed using machine learning algorithms via advanced techniques such as classification algorithms like Random Forest Classifier (RFC), AdaBoost Classifier with a Decision Tree (DT), Logistic Regression (LR), -Nearest Neighbors (KNN), Gaussian Naive Bayes (GNB), Decision Tree Classifier (DTC), Support Vector Machine (SVM) to analyze various features and patterns within the data. The results show that the AdaBoost with a Decision Tree model did well using a variety of assessment indicators, including Accuracy, Precision, Recall, and F1-score; the model aims to classify customers, indicating their likelihood of raising complaints in the future. The results and insights obtained from this study can provide telecom service providers with valuable guidance for implementing data-driven strategies to improve customer experience and take proactive maintenance that will reduce cost and churn.