Stock price forecast using clustering and meta-initiative algorithm (Neural networks )

In this study, arguing that both technical analysis and prediction using classical methods have a high potential for predicting price and stock behavior in the future; a model has been developed in the form of a wide-ranging neural network for predicting stock prices.For this purpose, at first, three databases were constructed to forecast the price of three shares of Parsian Bank, Mobarakeh Steel, Esfahan and Fars chemical industries. Then, according to the clustering method, the row and column headings of the databases were reduced and the most similar time series to the target time series were identified in each database. Finally, the remaining criteria in the databases were entered as neural network inputs with a hidden layer and the network learning algorithm was selected.The results of the proposed network were compared with the predicted results by multiple regression and time series regression methods and two criteria of MSE and MAD were investigated. The function of the neural network was well-seen in predicting prices in all three databases and both.

Keywords:

Hierarchical Clustering , Meta - initiative algorithm , Neural network , Stock Price Prediction