Prıce-Predıctıon Analysıs Usıng Neural Networks
In this study with the aid of artificial neural networks using price history information of products, we have tried to predict price changes in the upcoming time steps.
First of all we will have a brief definition of “prediction” and then artificial neural networks are mentioned as a suitable tool for prediction due to their learning abilities and other specifications. Subsequently artificial neural network types which are generally used and suitable for prediction are briefly discussed.
Subsequently earlier studies on price prediction using neural networks with an implemented model is discussed. And also an alternative model that we tested for this project but was not successful is included. The disadvantages and weak points of these models are listed afterwards.
In the section 3 the method used in our project is described. First price history data format and data preparation process for usage is explained. Subsequently Narx Artificial neural network model is briefly interduced. Then the artificial neural network model designed for this project and its input, output and system logic is described.
Subsequently the project implementation in Matlab environment is explained with details. Used database table formats and general program flow takes place in this part.
In the section 4, the system results generated for thirty different products fetched from database, is shown in figures. Subsequently conclusion and system performance analysis is done. In addition for testing that system generates logical outputs, system result for assumed price history in easily guessable sinus and cosinus function formats are displayed in figures which meet our expectations.
Dostları ilə paylaş: | 
