Research On The Investment Decision Of Chinese Financial Market Based On Chaos Theory

In Recent years, as the organic elements of market economy, the global financial market growsrapidly, and become more and more important. Chinese financial market, as an emerging market, itsgrowth has caught the attention of the world. The China A-share market value has jumped to thesecond place in the world. Chinese futures market has boomed in recent ten years and become thelargest commodity market in the world. The Shanghai and Shenzhen300index futures, the firstfinancial futures contact, have listed in2010. Along with the awakening of the domestic investors’risk aversion; Chinese gold market has led to enormous progress in trading volume and market impactdespite its late start. Financial investments gradually becoming important financial tools forindividuals, business, and government in China.In financial analysis and investment decision fields, Efficient Market Hypothesis (EMH) andCapital Asset Pricing Model (CAPM) are headstones at all times. As the time develops, people foundgradually that financial market is a complex system which is fractal and chaotic. Based on the chaostheory, Chinese stock, futures and gold market are studied systematically in this thesis. The purpose isto discover the inner discipline of these emerging markets, and to discuss the method of makinginvestment decisions.The main research contents include four parts:1. Chaos criterion for Chinese financial markets. In data pretreatment, two de-trending methods ofRate of Return (ROR) and Log Linear De-trending (LLD) are taken. For futures contacts’ priceswhich are discontinuous, a Max-Volume Restoration (MVR) method is designed to meet therequirement of price continuity and representation. Then the nonlinearity and determinism isinvestigated by using the method of R/S, BDS, and recurrence plots. Further more, by using the PhaseSpace Reconstruction Technique (PSRT) method, the chaos invariants are calculated. The analysisoffsets the deficiency that chaos has not been detected in not only most of domestic futures contacts,but also in gold market. At last, the conclusion is drown that chaos generally exists in Chinesefinancial markets.2. Studying on the noise processing of Chinese financial market. The thesis researched the noiseprocessing from the following two aspects: one is the noise estimation. The noise level of Chinesefinancial market is estimated by using some common method such as correlation integration method,coarse-gained entropy method and wavelet method. We use the function of wavelet variancedecomposition to analyze white noise, found a new method of noise estimation. The other is noise reduction. We analyzed nonlinear local average method and local geometric projection, focused on thewavelet soft thresholding method. A new method of noise reduction by soft-thresholding is found andtested in Lorenz and Chen’s system. Then we show it is effective while de-noising price series ofsome representative variants from Chinese financial market. At last, we take Shanghai StockExchange Composite index (SSEC) as the sample, through one-day-forecasting, de-stationary, andcomparing the Root-Mean-Square-Error (RMSE), to compare the actual de-noising effects of thosemethods in Chinese financial market.3. Studying on the chaotic predictability of Chinese financial market. First, based on noiseestimation and reduction, some representative variants from Chinese financial market aredemonstration forecasted with maximum Lyapunov exponent method. Then the Valterra seriesadaptive prediction in Chinese financial market is studied. RLS algorithm is proposed to improveprediction accuracy. The result shows that the Valterra series adaptive prediction method issignificantly better than maximum Lyapunov exponent method. Generally, commonly used neuralnetwork models belong to static feed forward processing mode. In this thesis, the Recurrent PredictorNeural Network (RPNN) is applied to the prediction of Chinese financial market. We train thenetwork by using Genetic Algorithm (GA) to optimize the weight and threshold, also the amplitudeand slope of the excitation function. The improved RPNN is compared to two classic neural networks–Back Propagation Neural Network (BPNN) and Radial Basis Function Neural Network (RBFNN).The result shows that the proposed method is more accurate and stable, and is an effective predictionmethod for Chinese financial market.4. Studying on the chaotic trading model and portfolio model of Chinese financial market.Technical analysis is the most widely used analysis tools for financial investment decision. From theperspective of chaos and fractal, the three assumptions of technical analysis are reinterpreted. Thethesis considered that the development of chaos and fractal theory has strengthened the theoreticalfoundation for technical analysis. Then we combined chaotic forecasting with technical analysismodel, created and tested some chaotic trading models, include moving average rules and Alexander’sfilter. The mixed trading model is built by combining chaotic forecasting, based on Genetic Program(GP). The model is tested in Chinese financial market. The result shows that it is better thantraditional trading model in both excess return and stability. At last, based on nonlinear and behaviorfinance theory, a utility function–skewness portfolio model is formed based on loss aversion, and itoutperforms traditional portfolio models.