Application of hierarchical time series model with transfer function

The prediction of the amount of water needed in each region is important since clean water is currently needed by every community worldwide. The monthly water consumption depends on how many users. The users vary from people, industry, agriculture, etc. Water providers, in this case local governments, need to estimate the needs in the short- and long-term periods. In the long run, water demand prediction is beneficial for policymakers as the population and water users are increasing such that efforts are needed to supply local water needs. In this study, we aim at predicting the need for monthly water consumption based on users or customers in the city of Malang, East Java, Indonesia and in each district within the city of Malang. There are five districts within this city. The water demand data are stratified from districts and city such that the modeling can accommodate hierarchy. Modeling and forecasting are done with ARIMA (Autoregressive and Moving Average) and transfer function models to accommodate one or several variables affecting water consumption. Accuracy results (Symmetric Mean Absolute Percentage Error) indicate an accuracy error of approximately 3 percent. This result is quite satisfactory and can be used to estimate the accuracy of water demand in Malang city and its districts.