Area Grouping Based on Inflation During Covid-19 Using Minimum Message Length

Clustering in the form of Mixture Modelling has been widely studied to group data into clusters. Several methods have been used including Maximum Likelihood (ML), Information Criterion by Akaike (AIC), and Bayesian Information Criterion (BIC) by Schwarz. In this paper, Minimum Message Length (MML) is applied to the mixture modeling of inflation data. MML is used simultaneously for both parameter estimation and model selection processes. The aim of the study is to group areas in Indonesia, based on the inflation pattern during the pandemic Covid-19 period. The study investigated inflation time series data in the form of descriptive statistics and linear regression parameters as extracted features. Three settings are set in this study including the modelling using the extracted features of descriptive statistics, linear regression parameters, and the combination between the two. The results show that the three settings of modelling resulted in six, six, and four groups’ models respectively. One group had a specific inflation pattern with a high fluctuation. This group consisted of small number of areas such as Banda Aceh, Lhokseumawe, and Meulaboh, as well as Kota Bau-Bau, and Manokwari for one setting of modelling. All other resulting groups had similar features with differences in one or two other extracted features. Areas clustered in each group in the three settings of modelling are different between one group to another with only slight similarities of areas. The different settings of extracted features are the cause of this condition. However, in this study, it was also found that, for some areas, they shared a similar set of extracted features and inflation pattern between them during the pandemic Covid-19 period.