APPROPRIATE MODEL ANALYSIS FOR GARMENT SALES FORECASTING

Abstract

This paper demonstrated methods to create predictive models for garment sales. The
objective is to find the optimal model for garment sale forecasting. Dataset for creating
forecasting model is collected from daily garment sales of a well-known store from January 1,
2015 to September 30, 2017, totally 1,004 days. This work applied two main models. Those
are Time Series Analysis with decomposition and Regression Analysis. The second dataset,
which has 92 data points of daily garment sales from October 1 to December 31, 2017, is
employed to evaluate and compare the forecasting efficiency of both forecasting models. Five
deviation analysis methods are used to determine forecasting efficiency: 1. Basic Error 2.
Mean Absolute Deviation (MAD) 3. Mean Square Error (MSE), 4. Mean Absolute Percentage
Error (MAPE) and 5. Theil’s U. The study indicated that Regression Analysis gave more
accurate forecast than Time Series Analysis. Therefore, Regression Analysis forecasting
model is more appropriate to forecast daily garment sales for this store than Time Series
Analysis forecasting model.