International Journal of Engineering and Technology

Prediction of time series data, whether the data related to the stock market or exchange rate is of enormous interest to traders or investors due to high profit in trading. Thus the prediction of time series data indices and its study are important to discover whether the next day’s closing price would increase or decrease. For that reason, we construct a framework for predicting time series data by using a low complexity, adaptive functional link artificial neural network (FLANN) over a time frame varies from one day ahead to several weeks. The FLANN is basically a single layer structure in which non-linearity is introduced by enhancing the input pattern with nonlinear functional expansion. The architecture of FLANN will be trained with the working principle of heuristic stochastic algorithms to achieve the best forecasting and classification to increase in accuracy of prediction and decrease in training time. Heuristic and Stochastic refers to an experience-based technique for problem solving, learning, and discovery that find a solution which is not guaranteed to be optimal, but good enough for a given set of goals and helps in minimizing errors. The proposed project is based on all these parameters which will predict time series data using FLANN trained by heuristic stochastic algorithm. Extensive computer simulations will be carried out and it is compared with traditional artificial neural network to observe the accuracy of the FLANN model.