Computer Science, Machine Learning and Big Data

Forecasting is necessary for better business understanding and decision making. When it is done manually, it requires a lot of effort and time from multiple departments like logistics, sales, finance, etc. It also involves lot of gut feeling from experienced people and sometimes it might lead to error prone prediction if person is inexperienced or is not aware of past behavior. It is even very challenging for any data scientist to find a forecast model that performs best for all scenarios and in all forecast horizons. In this paper an approach for forecasting using ensemble model is discussed. Ensembling is done using symmetric mean absolute percentage error and mean absolute percentage error calculated from rolling forecast approach. For validation of forecast model, M3 competition data is used. This approach has resulted in better performance on out of sample prediction.