UTILIZATION OF MACHINE LEARNING ALGORITHMS FOR CUSTOMER COMPLAINT CLASSIFICATION AT PERUMDAM

Abstract

In today's digital era, managing customer complaints poses a significant challenge for public service providers, such as the Regional Drinking Water Company (PERUMDAM). With the increasing number of customers and the complexity of complaints, manual methods are no longer adequate to efficiently handle large volumes of data. This study focuses on the application of machine learning algorithms, namely Logistic Regression, LightGBM, and Random Forest, to classify customer complaints based on patterns in the data. The research stages include dataset collection, data preprocessing such as cleaning, casefolding, tokenization, filtering, normalization, and stemming, as well as feature extraction using TF-IDF, SMOTE features, and hyperparameter tuning. Evaluation is conducted based on metrics such as accuracy, precision, recall, and F1-score. The results indicate that the LightGBM algorithm provides the best performance in terms of precision and recall, while Random Forest achieves the highest accuracy, and Logistic Regression serves as an efficient model for data with linear relationships. Therefore, LightGBM is recommended for further implementation in managing customer complaints data quickly and accurately. This study contributes to the development of technology-based solutions to enhance the efficiency of customer complaint management in the public service sector.