Research And Application Of Securities Time Series Forecasting Model Based On HMM - ANN Hybrid Model

Currently, there are many popular research methods for the prediction of time series, such as Statistical method, computer science algorithms, artificial intelligence, mixed model methods, etc. All kinds of scientific methods have their scopes, so when we study some forecasting methods, e should fully consider about the characteristics of the subjects, and then improve methods according the internal and external factors so that we could achieve the purposes of accurate prediction. In the paper, e select the securities herein as the experimental time series data and research object, and the hybrid model as the study of methods, taking into account the number of the main factors affecting the Chinese stock market are large, such as multi-market, the policy, the technology, the finance, he fundamentals and other factors, etc. Every aspect contains its specific content, therefore, when we predict China’s stock market rise or fall, we can not lump under one head.Based on the existing model in the various research areas, this paper selected the stock which index from securities market and try to consist of sequence databases, which we can use the model of this article to predict more accurately. On the forecasting process, the hidden Markov model, BP artificial neural network model, regularization method, genetic algorithms to integrate together to form intelligent organism for data processing.According to the experimental data, the paper made quantization and screen in the selection of store and prediction of index. After divided the industry sectors of the stock market, we can find the highest accuracy for each index, the steel plate, in the formation of34types of trades. For the steel plate, we selected the index from the higher forecast earnings to predict. Finally, the revenue for each share in the steel plate has been defined as the object of model study,and has been compared with other three typical models on error analysis methods, thus it confirmed the superiority and effectiveness of the model in the paper.