Machine Learning Approaches for Analyzing Bank Customer Complaints

Proper customer service is the key to increasing customer retention and brand credibility in financial banks. This can be achieved by helping solve customers” issues through their complaints. This study explores the application of traditional machine learning models to classify bank customer complaints. Using the dataset from the CFPB database, this research implements four ML models to handle this task: Multinomial Naïve Bayes, Random Forest Classifier, Stochastic Gradient Descent Classifier, and Logistic Regression. The dataset first underwent preprocessing steps to clean irrelevant data using Natural Language Processing and the features were extracted using Term Frequency-Inverse Document Frequency. After evaluating the baseline results, hyperparameter tuning is implemented using GridSearchCV to analyze improvements in performance for each model. The results show that LGR achieves the best performance out of all, with the highest accuracy score: 0.827 and an F1-score: 0.825. In conclusion, traditional machine learning models have the potential to provide a cost-efficient solution for classifying bank customer complaints.