Comparison of Three Time Series Forecasting Methods on Linear Regression, Exponential Smoothing and Weighted Moving Average

Abstract

The purpose of this study is to compare the 3 forecasting methods Linear Regression, Exponential Smoothing and Weighted Moving Average based on the smallest error value or close to zero. From the results of this study, the Linear Regression method was obtained as the correct method with a predicted value of 502 students, the smallest error value was MAD 27.83, MSE 1152.1 MAPE 8.1%. The Tracking Signal value moves between 1 and -1, the movement is within the control limits of the tracking signal standard deviation distribution 4 and -4, meaning that the method is correct. The Moving Range value moves between 68 and -46, this value is within the MR control limits of 117.83 and -117.83, this result shows that this means that this method has been tested for truth and can be accepted as well. Thus, indicating that the Linear Regression method as a forecasting method is appropriate and acceptable as a basis for future decision making. The level of accuracy of the error and the value in control shows that there is a time series data relationship between the x variable, namely time, and the variable y, namely actual data. In addition, it produces trending data movement patterns, meaning that data movements experience a significant increase over a long period of time or for 7 periods.