<![CDATA[Rural cooking oil price data was collected in 32 provinces in Indonesia. Forecast model can be in form of ARIMA for each province. Forecast is used to formulate precise controlling policies of rural inflation. Time series data can have pattern resemblance. This resemblance can be used to form clusters. Therefore, this research aimed to compare the accuracies on ARIMA model of cooking oil price developed using hierarchical agglomerative clustering (HAC) time series method which its forecast is used by each member of clusters. As the results, modeling for 32 provinces individually resulting an average MAPE of 2.58%, while ARIMA model development with CID as the distance measures for HAC and employing 2 or 5 clusters, gave a more efficient and speed on modeling process and only giving effect on increased MAPE just by average of 7%, furthermore when 5 clusters was introduced, it resulting an increased of MAPE just by 4%. These were all compared with models without clustering. Those clusters with CID gave a lower MAPE in 10%. The CID cophenetic coefficient was 0.83.]]>
Copyrights © 2021
