Unsupervised Learning Approach for Evaluating the Impact of COVID-19 on Economic Growth in Indonesia

The spread of COVID-19 that occurred in several parts of Indonesia resulted in the economy getting worse. Almost all business fields in Indonesia experienced a contraction. Each province has a different impact from one another so that the policies taken cannot be generalized. Therefore, this study was conducted to group provinces based on the value of Gross Regional Domestic Product (GRDP) in 2019 and 2020 using unsupervised learning. The year 2019 represents the economic conditions before COVID-19 and 2020 represents the conditions during the COVID-19 pandemic. The unsupervised learning method used in this research is the K-Means, K-Medoids, SOM, as well as Hybrid SOM-K-Means methods. From the grouping results obtained, then a comparison of the results of the four methods will be carried out to obtain the best method based on the Silhouette Index. The results show that based on 2019 data, the grouping of provinces using the K-Medoids, SOM and hybrid SOM-K-Means methods is three clusters. While the results of grouping using the K-Means method are as many as two clusters. On the other hand, based on 2020 data, both the K-mean, K-Medoids, SOM, and hybrid SOM-K-Means methods show the same results, namely the grouping is carried out in two clusters. The best method based on 2019 GRDP data is K-Means with two groups. Meanwhile, in 2020, the best method obtained is the K-mean, K-Medoids, and SOM methods with two groups. In addition, all economic growth indicators have contracted or decreased due to the impact of the COVID-19 pandemic.