Financial Risks Prediction Of Time Series

In this paper,starting from the harmfulness of the financial crisis,focus on the relevant parameters affecting the financial risk is more noticeable,building model,predict the impact on our country risk situation.We can effectively find the potential financial risks and to serve the macro aspects of the financial market operation,improve the accuracy of the financial regulation.The article first analyzes the financial risk prediction research in the field at home and abroad,noted that China's financial system risk prediction model construction and related research aspects of the status,for the current domestic financial early warning model at an early stage,using data mining techniques FCM clustering and wavelet contrast clustering algorithm,clustering algorithm wavelet addition to classification characteristic that also has the characteristics of mufti-resolution wavelet,and the advantages of filtering denouncing.As the financial data sequence having a non-linear characteristics,we introduced the chaos theory of phase space reconstruction prediction model in forecasting models used RBF neural network model as a fitting function.We will then wavelet fusion algorithm clustering and chaotic use both clustering algorithm to achieve the simple neural network structure,the adjustment parameters related neurons,reducing redundant information neural networks,to prevent over-fitting,and ultimately improve forecast performance purposes.Finally,we test certification,the financial risk prediction chart can be seen clustering wavelet chaotic prediction accuracy has been significantly improved,the prediction of the data,access to financial early warning,as shown by the warning to guide decision makers to take appropriate measures.According to in-depth study of the pr-wavelet algorithm and chaotic phase space reconstruction of the local model,understand broad application of these algorithms,these algorithms are applied to turn the financial sector.By reading a lot of literature,hybrid algorithm can understand the disadvantages,and play the advantages of individual algorithms,so I decided to join wavelet clustering or FCM clustering algorithm and chaotic time to go to the local model,and through empirical to verify,the empirical results show that the hybrid algorithm has better prediction effect and higher accuracy,and it can meet the requirements of the application...