Improved Rough-Multiple Regression for Unemployment Rate Model in Indonesia

Many statistical approaches have been employed to model unemployment data such multiple regression and time series approaches. However, most researchers are not really concerned with inconsistent information in the data set before going to model building. While the accuracy of model prediction depends on data quality or data input. In this paper, we adopt rough set theory into multiple regression model for investigating the inconsistent sample and variables, respectively. This adoption is also considered to reduce a large number of independent variables. Some modifiable variables are related to unemployment rate, including labor force participation rate, gross domestic product, human development index, number of population and province in Indonesia. Interestingly, there are 34 provinces as dummy variables in representing of the geographical variable. The main objective is to measure and compare relationships between unemployment rate with these modifiable variables before (2019) and during (2020) in the pandemic era based on conventional multiple regression and proposed rough-multiple regression models. All information regarding variables above were gathered from Statistics Indonesia. The results showed the proposed rough-multiple regression able to improve the coefficient determination significantly from 67.6% to 85.9% for 2019 and 67.3% to 84.5% for 2020, respectively. While no significance different of this coefficient if derived with conventional multiple regression model. The proposed model better than conventional regression in explaining of the variation of the unemployment rate in Indonesia.