Industry Matching Quantitative Investment Strategy Based On Neural Network Technology Research

Along with the continuous evolution of China as an economy, the role of financial markets in the process of resource allocation has become more prominent than ever. In an era witnessing advancements in the globalized markets and a group of innovations in financial instruments, a stable financial market has become the dominant factor in securing the smooth development of an economy, generally reflecting a country’s prospects.In the progressive development of the stock market in China, the superseding rises and falls among various industries has been attracting investors’ attention. By tracing the rise/fall patterns, prediction of the industry parties in terms of their relative strength is feasible. This may serve to reveal the law of the operation of stock markets, as well as to guide the investment practice.Due to such features as nonlinear mapping and adaptive sampling, the neural network scheme has been drawing of attention since its invention. Following its successes in assorted engineering discipline, e.g., patter recognition, intelligent control, etc., researchers began to recognizing its ability in extracting data’s complex nonlinear relationship as a hopeful tool in uncovering the relationship among variables in financial markets.The first part of the thesis analysis the stationarity and cointegration of PE ratio series in assorted industries, with a conclusion that the difference in the series of PE ratio tend to present a mean-reverting feature, rendering the relative valuation among industry parties to be more stable than the absolute valuation of a single industry alone. The second part introduces the BP neutral network into our study. Furthermore, in accordance with the forecast subject, we optimized the input variables, the network structures, the initial weights, and the training methods. Finally, taking financial service industry and bio-tech industry as an example pair, the thesis made predictions on the relative strength between the two in the simulation period of2011.1.4-2012.2.29with the rolling method. Through the simulations, we evaluated the feasibility of the BP-neutral-network-based prediction of the relative strength among industry parties, and therefore proved its significance in practical investment guidance.