An Integrated Model for Time Series Price Prediction Based on MODWT and BP Neural Network

Time series exist widely in the real world and are often used to study price forecasting. By modeling the time series, it is possible to predict the future values of specific research objectives. With the continuous improvement of people’s income and financial literacy, they are increasingly emphasizing the mining and analysis of data in the financial market. Therefore, the stock price time series is often used for quantitative investment modeling to identify investment opportunities. However, the time series in the financial market is relatively complex, and the relevant data has characteristics such as nonlinearity and non-stationary. These characteristics make it difficult to achieve research objectives using general statistical econometric methods. While machine learning methods can effectively process complex data. In fact, with the continuous development of machine learning technology, it has also been used for related research in the financial market. Considering this, this paper uses BP neural network to model the time series. Furthermore, considering that the original time series and the new subsequences obtained by decomposing the original time series based on specific techniques have different characteristics. By studying new subsequences and modeling them, better research results can be achieved. Therefore, this paper further constructs an integrated model based on MODWT and BP neural network to forecast the time series price. It is found that the integrated model based on MODWT and BP neural network shows better prediction effect, with smaller root mean square error and mean absolute error, and the fluctuation range of error is smaller and more concentrated.