Comprehensive scoring trading model based on LSTM prediction

This paper studies a trading model for short-term trading strategies, through which maximum returns can be achieved while maintaining low risk. First filtered through to the data, we found that there are 10 days missing value in gold trading data, in order not to affect the data calculation and analysis, we will be ten days as for some special reasons that led to a gold cannot trade, after eliminating the 10 data, gold has a 1245 - day transaction data, the currency has 1826 days transaction data. Then we made predictions about the data. After compared with other models to predict the effect, take the LSTM both short-term and long-term memory network learning, according to the characteristics of the gold and BTC make different learning date, eventually won the gold and BTC price estimate, The correlation coefficients between the predicted price and the real value are all above 0.95, and the trend is very appropriate and the effect is excellent.