Prediction for time series by using fuzzy nearest neighbor method with application

In this research we study the fuzzy nearest neighbor method (FNNM) for time-series prediction , this method depends on fuzzy membership value .The main goal of the prediction algorithm is to forecast future value depending on past nearest neighbors value. The nearest neighbors that we choose by using fuzzy membership value or portmanteau membership value .To measures the accuracy of our method and to compare with ARIMA model we use mean absolute percentage error (MAPE) and mean square error (MSE) that is calculated from the actual value of time series data and the number of internet users and forecasting value. The results encourage using fuzzy nearest neighbor in forecasting.