Open AccessAnalysis of Multivariate Adaptive Regression Spline (MARS) Model in Classifying factors affecting on Student the Study Period at FKIP Darussalam University of Ambon
Author(s) -
Safarin Zurimi,
Darwin Darwin
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1463/1/012005
Subject(s) - mars exploration program , multivariate adaptive regression splines , statistics , multivariate statistics , mathematics , ordinary least squares , regression analysis , function (biology) , spline (mechanical) , value (mathematics) , regression , computer science , bayesian multivariate linear regression , engineering , biology , evolutionary biology , astrobiology , structural engineering
This study aims to determine the applicative model that can be revealed through the Multivariate Adaptive Regression Spline (MARS) classification Model for the problems alumni study period of the FKIP Darussalam University of Ambon. To find out the applicative model that can be revealed through the MARS classification model, parameter estimation is first performed to find the best MARS model. The best MARS model is chosen based on the minimum generalized cross validation (GCV) value.The research result indicates that the parameter estimation of MARS model using Ordinary Least Square (OLS) method obtained were convergent. This is indicated by the smallest of value Mean Square Error (MSE). In addition, the application of the MARS model using the OLS method on data alumni study period of the FKIP Darussalam University of Ambon, result indicates that there is an influential base function, BF3 which contains two predictor variable, that is first semester achievement index and family economics conditions. This is caused by BF3 has a significance value t count > t table so the decision taken is to reject H 0 which means that the BF3 basis function parameter has a significant effect on the model.