Development of Machine Learning Based Price Quote Prediction Models for Electronic Payment and Money Institutions

Nowadays, the economic world is constantly changing. Electronic money institutions that provide electronic payment services must constantly adapt to this changing market. These institutions need to set new commission rates in accordance with this market, which can be a time-consuming process. Incorrect commission amounts can sometimes even lead to customer loss. In this context, electronic money institutions need to take some strategic actions to increase customer loyalty. Price quote prediction stands out as an important strategic action that helps determine commission rates. This study aims to develop prediction models to help electronic money institutions determine accurate commission rates and, in turn, maximize customer satisfaction. For this purpose, the “Single Withdrawal Commission Rate” value has been predicted by using Extreme Gradient Boosting (XGBoost) and Extreme Learning Machine (ELM) methods. The forecast models have been trained on a dataset containing 1,707 rows of data collected from Moka United between February 1, 2021, and September 12, 2024. Additionally, the effect of $\mathbf{F}$-Regression in feature selection has been examined. The performance of the models has been evaluated with Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) metrics. The findings suggest that the incorporation of F-Regression contributes positively to the performance enhancement of the developed models.