Chotic-Recurrent-Neural-Netowrks-for-Financial-Forecast

We conducted the forecast experiment for our algorithm in various financial index datasets from 2010 to 2020. All datasets utilized in experiments are fetched from the free and open financial data community Tushare, a professional financial data sharing platform. In our experiment, eight datasets from two categories are used, including 1) Financial index: Dow Jones Index (DJI), Hang Seng Index (HSI), Nasdaq Index (IXIC), S&P 500 Index (SPX), Shanghai Securities Composite Index (SSE), Compositional Index of Shenzhen Stock Market (SZSE); 2) the stock of Apple Inc. We downloaded the raw time series data, around 10 years, and we split the 70% front of data as the training set, the rest data was utilized as out-of-sample verification. The forecast object is the clos-ing price of each dataset, using 30 days closing price to predict the 31th closing price. The test was conducted from the perspective of closing price forecast error which in-volved three indicators, such as 1) mean squared error; 2) root mean squared error; 3) mean absolute error. We conducted the experiment in the Python 3.8, relying on the packages of Tushare, Numpy and Pandas. In the model construction, we developed the CRNN and its comparison model FFBPN, RNN, LSTM with Pytorch which is a power-ful machine learning framework for neural-network-based academic research and de-velopment. Moreover, we trained all algorithm with 100 epochs using the same Opti-mizer and Loss Function, ADAM (Adaptive Moment Estimation) optimizer with learning rate 0.001 and MSE (Mean square error) Loss Function. All the models in-cluding CRNN and its comparison model were running in one hardware computing environment: Win 10 operating system, Intel(R) Core (TM) i5-8265U CPU @ 1.60GHz, 8g memory.