Operations Research on Customer Complaints in Fintech using ARIMA & Holt-Winters Method and LSTM (Long Short-Term Memory Networks)

Abstract

This paper aims to analyse and forecast customer complaints for the fintech industry based on ARIMA, Holt-Winters, and Long Short-Term Memory (LSTM) networks. Given the rise of the use of digital financial services, prediction of complaints enables firms to improve customer relations and system performance. The study utilizes only secondary research data, newspaper articles, journal articles, case studies and other industry-available sources. In comparing the two methods, it reveals that one technique has some advantages over the other while the other has limitations. Some research has demonstrated that using both forms enhances accuracy in prediction. It is useful to identify some guidelines and recommendations for fintech firms to apply to minimize the number of complaints received concerning customer service.