Neuro fuzzy logic approach with forecast modelling: The case of Turkey for unemployment rate

The growing interest of forecast modelling has brought the variety of models in order to plan and control the future. Neural Network, Fuzzy Logic and Neuro Fuzzy Inference Systems, which have significant places inside the artificial intelligence studies that are alternatives against both cause and time-based forecasting studies, have been used in remarkable researches. In this study, neural network, the combination of neural network and fuzzy logic technique, which is a hybrid technique, is called Adaptive Neuro Fuzzy Inference System (ANFIS), are used. The purpose is to determine the best technique which will forecast to reach the accurate and reliable results among them. However, it has been decided not to use frequently used variable. The aim in this point is to decide the best technique while forecasting the new variable instead of common used variables and techniques in literature. Therefore; used variable, known to be important for businesses and macroeconomic, is determined as unemployment rate that is not frequently used variable in researches. As a result, in this study, monthly unemployment rate data between the years of 2000-2012 taken from Turkey Statistical Institute are used to forecast monthly unemployment rate of 2013 for Turkey. The best model design for each forecasting model is selected among time-based forecasting models which are built by using both neural network and adaptive neuro fuzzy inference system (ANFIS) and also first of all, the first 6 months of 2013 is forecasted by using this selected model. In order to increase the reliability of forecasting result, the actual values of the first 6 months of 2013 has been compared with the forecasted values of the first 6 months, and with regard to statistical view, ANFIS is determined to have better forecasting performances than neural network has. In this manner, the second 6 months of 2013 are forecasted by using ANFIS technique. Neural network and ANFIS applications are performed using MATLAB.