A New Smart Chaotic and Noise Based Approach to Predict Tehran Stock Prices

mohamad hosein panahi¹ dr. babak karasfi²

1) researchers
2) supervisor

Publication : 3rd international conference of Science and Engineering(3icesconf.com)
Abstract :
Stock is the mirror of economics country . To forecast the marketing in different countries there are different Techniques , but the computers make an revolution on Trend Forecast of marketing and the Methods of auto regressive and mathematics is applied on the forecast. By the cause of non linear of this series didn t give a good result and non linear type is replaced. neural network is one of the trend forecast finance time series is one of the important tool .There are many researches for Prediction of stock index. In IRAN researchers for Prediction of stock index is offered differently regresoni linear method and mathematics modeling or even non linear method Including normal neural networks , fuzzy, genetic,But these methods are unusual Circumstances Exchange and chaotics are not affected in their forecasting.On the researches out of IRAN Many of the world s leading stock including Netherlands, China, Taiwan, India, Turkey. Mostly neural network method is forecasted. Research of foreign researchers in this base has continued until recently and it forecasts with High accuracy , There is still space for growth and improvement and using of analysis and Good algorithm helps to improve this research. This research seeks to offer a smart model of Expected prices in the Tehran stock market and comparing with other methods and Detecting of chaos in data and presentation Finding a way to eliminate it negative effects on forecasting and helping to improve accuracy. With giving attention to the Chaos are ignored in data To identifying non linear stocks and designing an occasion method for forecasting. Since that establishing a solution for eliminating these factors to polishing new method on modeling. At the beginning of making this model is forecasting the series of time forecast is considered and then non linearity of it and the chaos on this series are proved. The next step is forecasting on different models and it starts with the linear an then we go to normal neural network and trainin different structures of layers with repetition on data will be done that shows behavior of stock index is learned but still Chaos not resolved correctly. Following our discussion increasing in forecast accuracy, Imperialist competition Algorithm and Radial Circuit Functions is Used and tested. At each stage of improvements in MSE forecast error is Observed But at the end of algorithm RBF Standard error MSE Is improved The result of a combination of different methods In other related research is much more acceptable and Higher accuracy in predicting on forecasting.
Keywords : Stock Price Prediction Time series Chaos Neural Network Imperialist competition algorithm NNRBF