The Application of a Combination Model Based on Machine Learning in Offshore RMB Exchange Rate Forecasting

Abstract

With the vigorous development of China's economy, the internationalization level of Renminbi (RMB) is still improving, and the offshore RMB market is also expanding. Compared with onshore RMB, the price of offshore RMB is formed by free bidding of buyers and sellers. Market supply and demand are two extremely important factors affecting the exchange rate of offshore RMB, and the fluctuation of exchange rate is relatively free, so the offshore RMB exchange rate is relatively volatile. Under such circumstances, it is of great significance to forecast the offshore RMB exchange rate. This review focuses on the application of combinatorial models related to machine learning in offshore RMB exchange rate forecasting, and systematically combs the data types, modeling methods, and related forecasting performance involved in recent research. This paper evaluates the application effectiveness of the machine learning fusion Autoregressive Integrated Moving Average (ARIMA) model, supervised learning based on neural network and random forest, and the Long Short-Term Memory-Temporal Convolutional Network-Convolutional Neural Network (LSTM-TCN-CNN) hybrid model in offshore RMB exchange rate forecasting, compares and analyzes the advantages of each model, and summarizes the applicable scenarios of different models. The results show that the machine learning fusion ARIMA model is suitable for the prediction of time series with obvious rules and a small amount of data, the combination model of random forest and neural network is suitable for the prediction of structured data, and the LSTM-TCN-CNN model is suitable for the prediction of multivariate time series with a large amount of data and high dimension.