Stock Forecasting using Combination of Continuous Wavelet Transform and Neuro-Fuzzy Systems

Stock market forecasting is an important challenge for clients, so it attracts researchers to introduce new and reliable approaches to forecast this market. There are many approaches using wavelet decomposition to access details and approximation of stock market data and on the other hand, using neuro-fuzzy systems can be useful to overcome the uncertainty of data for forecasting. In this study, a hybrid approach will be introduced that is a combination of continues wavelet decomposition and neuro-fuzzy system to predict stock market data. For this aim, first data is normalized between 0 to 1 and decomposed by continues wavelet decomposition and data of each sample time was used to forecast “n” sample forward and finally, this approach was evaluated. For evaluating the performance of this approach, index of petroleum products of Tehran stock market between March 2015 to September 2019 was used and five steps of forecasting were considered. The best performance was 6.64x10-5 for one step forward forecasting and the worst was 4.74x10-4 for five steps forward forecasting. Results showed that, increment of forecasting step causes decrement of forecasting precision however it is reliable.