66 MODELING THE LEVEL OF OPEN UNEMPLOYMENT IN CENTRAL JAVA WITH MULTIVARIATE ADAPTIVE REGRESSION SPLINES (MARS) APPROACH Erma Oktania Permatasari 1 , Firda Nasuha 2 , Carlos L. Prawirosastro 3 1,2 Statistics, Institut Teknologi Sepuluh Nopember, Surabaya 60111, Indonesia 3 Universitas Hang Tuah, Surabaya 60111, Indonesia e-mail: erma.oktania@gmail.com ABSTRACT The level open unemployment is a value that shows the number of working-age population who looking for work, is preparing a business, feels impossible to get a job or already have a job but have not started working and often used for measured employment. Like at Central Java has increasing the total population at 2014 and have high total investation whereas should be can getting more employment, but actually still give high unemployment about 996.344 population at 2014. So that, in this research used nonparametric regression approach which multivariate adaptive regression splines (MARS) for modeling the level open unemployment in Central Java at 2014 because the level open unemployment in Central Java predicted influence by some factors. This research resulted in the best modeling for level of open unemployment in Central Java Province with value of GCV minimum that obtained at 0,396 with R-square at 86,5 percent as well as the predictor variables were entered into the model as much as three, namely the total population with interest rate of 100 percent, the minimum wage with interest rate of 41,955 percent, and the total working population with interest rate of 39,547 percent. Keywords: MARS, Nonparametric regression, Level of open unemployment 1. INTRODUCTION Indonesia as one of the developing countries have national development goals to be achieved, namely building a complete Indonesian people both materially and spiritually. One of the indicators used to assess the improvement of the quality of human resources is through work ethic possessed by each individual. But in this era of globalization, they are faced with a wide range of employment issues are characterized by employment growth rate is not comparable with the rate of growth of the labor force, low quality, competitiveness and labor productivity and investment climate was not conducive [1]. The highest number of populations in Indonesia is located on the island of Java. As is the case in Central Java, which increased total population of 258.324 inhabitants in 2014 with the highest contribution in Central Java is a textile investment [2]. But the problem today is that high total investment which can absorb a large amount of manpower, in fact they provide the number of high unemployment. The definition of the level open unemployment is a value that indicates the number of working-age population who are looking for work, or preparing for business, or find it impossible to get a job, or already have a job but have not started working. Previously had done some research on the level open unemployment rate. Among them are studies on the classification of districts / cities in East Java based the level open unemployment rate to approach MARS [3], the research on cluster analysis and discriminant analysis to classify the unemployment rate in the province of East Java [4], the research on modeling of unemployment in East Java uses regression spline [5], the research on modeling of unemployment in East Java using Fourier series [6], the research on the classification of unemployment based on the factors affecting the open unemployment provinces in North Sulawesi with methods CART [7], research on statistical analysis of unemployment rates East Java in 2012 [8] and the latest research on modeling the open unemployment rate in Central Java using regression spline [9]. Therefore, in this study will be used nonparametric methods, multivariate adaptive regression splines (MARS) to perform modeling of the level open unemployment in the districts / cities in Central Java in 2014 due to the level open unemployment in Central Java, allegedly influenced by several factors. In this case, it will also do a prediction based on the best model MARS obtained. International Journal of ASRO Volume 12, Number 01, pp. 66-74 Jan 2021