CNN-LSTM network fund price prediction method based on attention combination

Abstract

The invention discloses a CNN-LSTM network fund price prediction method based on attention combination. The method comprises the steps of 1, collecting fund platform data; 2, preprocessing the fund data; step 3, extracting sample features; step 4, establishing a fund price prediction network model; and step 5, training and predicting a fund prediction model. According to the CNN-LSTM network fundprice prediction method based on attention combination provided by the invention, the prediction precision is high, multi-step prediction of fund prices can be realized, and reliable reference information is provided for investors.

Description

technical field [0001] The invention belongs to the field of fund price prediction, in particular to a CNN-LSTM network fund price prediction method based on combined attention. Background technique [0002] With the development of the economy and the transformation of national concepts, fund investment has gradually become one of the important contents of people's life. It is very important for fund investors to analyze various factors that affect fund prices, grasp the changing rules of fund prices and predict prices, so as to effectively avoid price risks. However, fund price data is characterized by high noise, dynamics, complexity, multi-factor influence, and non-parameters. Obviously, accurate prediction of fund prices is a difficult problem with open challenges. [0003] With the improvement of financial theory and mathematical tools, various fund price prediction methods have emerged. [0004] Traditional fund price forecasting models use average line theory, K...

AI Technical Summary

Problems solved by technology

[0004] Traditional fund price forecasting models use average line theory, K-line diagram analysis, regression analysis, gray forecasting methods, ARIMA models, multi-core support vector machines and Markov chains, etc. However, since fund prices are affected by multiple factors, and these The factors show strong nonlinear characteristics, which makes the accuracy of these traditional linear fund forecasting models not high

In addition, most studies or methods are aimed at single-time-step forecasting of fund prices (that is, only predicting the fund price of the next day, month or year), which does not have much practical significance, because investors focus on a long period of time in the future price in period

Images