Research And Application Of Data Mining Methods For Singularity Characteristics Of Financial Time Series

The stock market is a complex system that is influenced by internal and external factors.It has many characteristics of interactive participation,and its market conditions are also ever-changing and intricate.The financial time series of the stock market is a real measurement and record of the historical data of the development of financial variables.It contains the inherent laws and behavioral characteristics of the financial market,so how to excavate more potential information and laws from it is extremely important for the investor's investment decisions and venture capital management.The singularity of financial time series plays an instructive role in investors' investment decisions.However,technical analysis models relying on mathematical modeling and statistical methods are difficult to be understood and mastered by investors because they involve multiple complex parameters.Therefore,in order to better guide investors on how to grasp the important opportunity for the stock to sell,buy or stop loss,this paper starts with finding meaningful patterns in time series to obtain implied feature information in financial time series.Although time series mining is currently a research hotspot,and many achievements have been made,theoretical methods and research for obtaining valuable information or meaningful fragments from financial time series are still lacking.So this paper focuses on the key issues that exist in it and applies data mining methods to research.The specific work is as follows:(1)Combining the influence of financial time series characteristics on mining,this paper de-noises the data by wavelet method.In this process,the selection of parameters such as wavelet soft threshold denoising process,threshold determination criteria,wavelet function and decomposition level are studied.(2)Combined with the k-line morphology theory,the feature sequences of different morphologies are detected by numerical calculation.The d-nearest neighbor clustering method based on wavelet transform is used to classify the feature sequences.The singularity of financial time series is analyzed longitudinally by experimental data.The degree of feature anomaly is of great significance to meaningful patterns of mining financial time series.(3)The significance of mining implicit patterns from time series is to hopefully obtain valuable information for future predictions.From the perspective of analyzing singularity characteristics of financial time series,this paper proposes a technical analysis model that can excavate the K-line Motif pattern and can proactively alert the stock price reversal trend.This model applies the motif theory to the K-line morphology theory and obtains the characteristic sequence of the K-line morphology.Then using data mining technology to classify the feature sequences,the statistical characteristics of the K-line Motif model are obtained.Finally,the K-line Motif model is used to verify the effect of the stock price reversal trend and the reversal point.(4)Finally,This paper empirically analyzes the historical trading data of Shanghai-Shenzhen A-shares.The results show that the K-line Motif model proposed in this paper does exist and can be well recognized,and the return on investment(ROI)of the short-term trend calculated based on the proposed K-line Motif model.it is proved that short-term trend theory applied to the time series to determine the impact of the trend of the stock is valid...