Equity Research Time Series Forecasting Model Based On Technical Analysis And CBR

Complex time series often contains a lot of potentially important information and rule, we analyze this kind of important complex data in order to reveal the inherent law of motion, change and development of time series. Finaneia time series is an important form of financial assets yield sequence, for example, the k-line indicates the price of stocks, funds, foreign exchange, financial derivatives and, etc, which is also the most important data in the field of economic and financial. Therefore, analyzing and forecasting such data in the financial investment forecast, decision-making and risk management is of great significance.At present, all kinds of market analysis techniques are applied to explain the stock market and forecast the future trend of the market. These technologies analysis not only require professional knowledge in finance and economics, but also need a large number of data collecting about the market, and a lot of calculations, on which the small and medium-sized investors have to spend too much energy. Human have robustness to solve problems, so the ability to handle problems will constantly strengthen with the growth of experience. Reusing previous experience method is a basic and important way for human experts to solve the problems. Because the CBR inference technology in artificial intelligence (AI) is very similar with human reasoning, so this paper put forward a stock time series prediction model based on technical analysis and CBR.Case reasoning (case-based reasoning, CBR) mainly consists four parts, which refers to case retrieval, case reuse and case modification and case save. At first, this paper applies the technical features of the given stock time series, and base on maximum minimum point to realize the pattern recognition of typical technology form, and also identify attribute value of patterns on start-stop time, volume, MA, OBV, RSI index. Besides, we take three consecutive morphological patterns as a complete case and build the case base. Then the similarity matching algorithm-NN case retrieval algorithm is adopted to search the target case from the case base to find whether similar case already exists, which can judge by the comparing similarity with threshold. Finally we realize the future prediction of time series of securities, and validate accuracy and effectiveness in theory and practical application of the model.