Pengelompokan dan Peramalan Deret Waktu pada Produksi Bawang Merah Tingkat Provinsi di Indonesia

Shallots are strategic vegetable commodity state that can affect the national economy. Shallots production increases every year that in line with domestic household consumption. Every province in Indonesia has a different level of shallot production, both in terms of cycles and harvest amount. Clustering provinces with similar production patterns can help government policies. This research aims to determine cluster time series and to evaluate the shallot production forecast in several provinces in Indonesia. There are three of optimal clusters which have a characteristic pattern in time series and their production. Time series at provincial level and cluster level, then it is modelled based on Autoregressive Integrated Moving Average (ARIMA) and Seasonal ARIMA (SARIMA). The evaluation of cluster level is compared to the provincial level and is concluded that clustering makes forecasting efficiently. This is based on average of Mean Absolute Percentage Error (MAPE) that is smaller that provincial level.