Mechanisms of the role of machine learning system patches in optimizing the performance of financial time series models

Abstract

As an important tool that can reflect the dynamics of financial markets, financial time series have their own characteristics and are susceptible to a variety of external factors, that will increase the challenge of financial time series analysis and modeling. The use of machine learning system patches in the era of big data can be achieved to optimize the performance of financial time series models and greatly improve the accuracy of financial time series forecasting. In this paper, the basic concepts and theories of financial time series, ARIMA model, LIST model and machine learning system are firstly analyzed in depth, on the basis of which a lot of problems existed in the financial time series model at the data level and algorithmic level are proposed. Then the structural performance of the two basic models and their shortcomings that require machine learning system patch optimization are analyzed. Finally, this paper analyzes the action mechanism of patch optimization from three aspects: data preprocessing, model training process, and model structure improvement, respectively. Through the analysis, it is hoped that the performance of the ascending financial time series model can be further improved to provide more possibilities for accurate analysis of financial market development dynamics.