Implementation of Light Gradient Boosting Machine (LGBM) for Customer Services Feedback Classification

Customer complaints are an important aspect in the business world. Every company must face complaints from its customers, whether related to unsatisfactory products, services or experiences. Customer complaints are a sign of unmet needs or expectations. A customer service person is tasked with serving questions asked by customers or consumers. The purpose of this research is to classify types of complaints to help customer service. The classification carried out is in binary form, namely disputed or not. This research compares several machine learning algorithms for classification of customer questions into appropriate categories for more efficient responses. Based on evaluation metrics, the F1 Score light gradient boosting machine (LGBM) is the best model compared to Multinomial Naive Bayes (MNB), Logistic Regression (LR), SVM (Support Vector Machine), and EXTC (Extra Tree Classifier).