Abstract
The Government are always making various efforts to establish sound financial management and stable budget policies.. In particular, the work related to the fund plan and the operation plan of the treasury through manual work is subject to improvement, and a financial estimation system is needed to achieve it.. In this study, we wanted to analyze whether it is possible to predict the balance of the treasury by utilizing machine learning. First of all, we developed a forecasting model for treasury balances. In addition, to verify the applicability, the government balance predictability of machine learning techniques, performance verification, and major variable verification were performed. In detail, data collection and refinement, exploratory analysis, analytical model design and development, and analysis result review were studied. The analysis target data is 120 months of total data from 2010 to 2019, limited to income tax and value-added tax revenue forecasts, with monthly revenue volume and major taxes. This was predicted for 2019 tax revenue respectively. The application algorithms are Linear Regression, Random Forest, and Gradient Boosting. According to the analysis, Linear Regression was the best in revenue and income taxes, while Gradient Boosting was the best in VAT. In conclusion, it was analyzed that machine learning was applicable to treasury predictions. This will allow quick results, scientific analysis through the introduction of big data-based models, and the person in charge will be able to improve predictive power through new model tests. It is also expected that the government
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call