Comparing SARIMA and Holt-Winters’ forecasting accuracy with respect to Indian motorcycle industry

Indian automotive industry is one of the largest in the world and has been growing at a very rapid pace. The industry is dominated by motorcycles with a market share of more than 70%. The Indian motorcycle industry has direct as well as indirect influence on the growth of the Indian economy, hence understanding and forecasting the performance of this industry is very critical. The key purpose of this journal paper is to compare the accuracy of Holt-Winters and Autoregressive integrated moving average (ARIMA) model which is popularly known as Box-Jenkins auto regressive model, in relation to Indian motorcycle industry and possibly suggest the best model. Currently, there are no studies exploring the forecasting accuracy, of two models with reference to Indian motorcycle sales. In this journal paper we equate the forecasted values of both the models and we choose the best model based on the least mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). Index Terms—Forecast, Holt-Winters, Autoregressive integrated moving average (ARIMA), Indian motorcycle, Mean absolute percentage error (MAPE), Mean absolute error (MAE), Mean square error (MSE), Society of Indian Automobile Manufacturers (SIAM). I. INTRODUCTION Planning is a crucial element in any industrial activity. However, for a business to be successful it requires goals that are built on sales projections, which intern depends on accurate demand forecast. Moreover, various other business assessments are too reliant on the prediction of future sales. Therefore, sales prediction becomes very critical for all planning and development activities. Accurate forecast helps companies to uphold ideal inventory level and sustain daily operations which in turn affect the profit margins. Therefore, companies and strategists benefit a lot from accurate demand forecast. Forecasters and statisticians use wide variety of forecasting techniques which differ in their complexity, amount of data required and ease of use (5). Out of these forecasting techniques, judgmental technique was found to be more dominant (9), (10). However, several studies have also shown that judgmental technique is less precise, prejudice and more likely to generate poor estimates than other methods (11), (12). Choosing suitable forecasting methods is very important to generate reliable and accurate forecasts (16), (17). However, each technique has its own limitation and fits only limited set of situations, thus forecaster have to choose techniques as per situations in order to get maximum accuracy. This journal paper thus efforts to weigh the two forecasting models and relate the outcomes and propose the best model for the Indian motorcycle industry.