Research On Predicting Method Of Interbank Offered Rate Trend

Shanghai Interbank Offered Rate(Shibor)is the benchmark interest rate for fund transactions between commercial banks with high credit rating in the financial money market.As the core interest rate of the interbank fund market,the interbank offered rate can comprehensively and timely reflect the quotation of the interbank offered rate for fund transactions in the interbank offered market.The trend changes of offered rate data affects the financial transactions in the interbank offered market,and then affects the risk management strategy of the central bank in the financial market.Therefore,it is of great significance to predict the influence of Shibor trend changes on the macro economy.The main research contents of this thesis are as follows:1.In the work of predicting the interbank offered rate data in the time dimension,first the differential autoregressive integrated moving average(ARIMA)model and the Prophet model were used to predict Shibor’s overnight varieties.Both of these two models are experimented in one-dimensional time dimension.Then,the grey model and the multivariate time series graph neural network(MTGNN)model are applied to predict the overnight varieties of Shibor from 16 quotation banks.The two models are tested in multidimensional time dimensions.Finally,the two small aspects of the work are compared experimentally.The results show that the Ptophet model is better than the ARIMA model in one-dimensional time dimension;the MTGNN is better than the grey model in multidimensional time dimension.2.In the work of predicting the interbank offered rate data based on the influencing factors of the characteristics of macroeconomic indicators,starting from the factors affecting the trend change of interest rates,first select nine relevant macroeconomic indicators and conduct correlation tests on them.Then select four indicators which are highly correlated with the change of the interbank rate and generate data for them,use the Time GAN model to convert the monthly data of these indicators into daily data.Finally put them into MTGNN and other models for experimental comparison.When dealing with the dimensional consistency of daily data and monthly data,use a suitable data generation algorithm to interpolate the monthly data to generate daily data,and compare the results of all prediction experiments.The experimental results show that the prediction performance of the experimental effect after adding influencing factors has not been improved too well,but the overall research and analysis show that when predicting the trend of the offered rate data,the prediction method used in this thesis is in the currency market.The practical application has certain guiding significance.