Comparison of forecasting performance based on time series estimation methods with randomness and trending data sets

Estimation has a vital important value in planning and directing the future and taking precaution against future unexpected situations. The success of the estimations made in a subject are as important as forecasting the future. Until today, numerous forecasting methods have been developed to be successful, namely, on the purpose of obtain high performance results. The estimation methods are basically divided into two grups that are based on time series and cause-effect relation. The scope of this study, Autoregressive Integrated Moving Average (ARIMA), Artificial Neural Networks (ANN) and Grey Prediction methods based on time series have been investigated to prediction performance in the time series of randomly dispersed and trendy data. So as to measure the estimation performance of aforomentioned methods, 64 different time series data set have been handled and their estimation performance were compared with Mean Absolute Percentage Error (MAPE) ve Root Mean Square Error (RMSE). As a result, though ANN present more suitable model for the data set, it has been deduced that Grey Estimation method makes better estimation in both random and trend data sets than other methods. Keywords: Time Series, Forecasting Methods, ARIMA, ANN, Grey Prediction