The article presents the results of a comparative analysis of two time series forecasting models — the classical ARIMA econometric approach and the XGBoost gradient boosting algorithm — applied to the exchange rate of the Chinese yuan against the Russian ruble. The role of forecasting models in algorithmic trading is described. The advantages and limitations of using machine learning methods in the foreign exchange market are considered. The data was interpolated, and the training and test sets were divided in a 80% to 20% ratio. The models were trained and tested, and accuracy metrics were calculated. All calculations were performed in the Python environment using the Pandas, NumPy, Matplotlib, Statsmodels, and Scikit-learn libraries. The results showed that XGBoost achieved a 24.4% reduction in mean absolute error compared to ARIMA, confirming its superior performance in limited sample sizes and non-stationary series.
exchange rate, ARIMA, XGBoost, algorithmic trading, time series forecasting, machine learning
1. Boks Dzh., Dzhenkins G. Analiz vremennyh ryadov: prognoz i upravlenie. — M.: Mir, 1974. — 406 s.
2. Dreyper N., Smit G. Prikladnoy regressionnyy analiz. — 3-e izd. — M.: Vil'yams, 2016. — 912 s.
3. Malyhin E.M. Algoritmicheskiy treyding dlya professionalov. — SPb.: BHV-Peterburg, 2021. — 176 s.
4. Oficial'nye kursy valyut Central'nogo banka Rossiyskoy Federacii [Elektronnyy resurs]. — URL: https://cbr.ru/currency_base/dynamics/?UniDbQuery.Posted=True&UniDbQuery.so=1&UniDbQuery.mode=1&UniDbQuery.date_req1=&UniDbQuery.date_req2=&UniDbQuery.VAL_NM_RQ=R01375&UniDbQuery.From=01.02.2026&UniDbQuery.To=30.04.2026 (data obrascheniya: 01.05.2026).
5. Harrison M. Mashinnoe obuchenie: karmannyy spravochnik. Kratkoe rukovodstvo po metodam strukturirovannogo mashinnogo obucheniya na Python. — SPb.: Dialektika, 2020. — 320 s.
6. Hesti T., Tibshirani R., Fridman Dzh. Elementy statisticheskogo obucheniya. — M.: DMK Press, 2020. — 764 s.
