| Summary: | Background: Classification method is a statistical method for grouping or classifying the systematically arranged data into a group so we can know that an individual are in a particular group. Multivariate Adaptive Regression Spline (MARS) introduced by (Friedman, 1991) is a methodology for approximating functions of many input variables given the value of the function at a collection of points in the input space. Although training times for this method tend to be much faster than feed forward neural networks using back propagation, it can still be fairly slow for large problems that require complex approximations (many units). Methods: This was a nonreactive study, which is a measurement which individuals surveyed did not realize that they are part of a study. Result: Based on the best model selection criteria MARS then the selected is with model BF 20, MI 1 and MO 0 with the form Y = 0.929944 + 0.912438 * BF1 - 0.218729 * BF2 + 0.886429 * BF3 + 0.215575 * BF4 + 0.0745423 * BF5 - 0.232014 * BF6 + 0.0472966 * BF7 - 0.0367996 * BF8 + 0.0188678 * BF9 + 0.0304537 * BF11. Accuracy of drugs user rehabilitation classification that non relapse and relapse status based on MARS model is calculated using precision classification value. The accuracy level of drugs user rehabilitation classification in East Java using MARS method produces accuracy of 95,71% and misclassification of 4,29%. The magnitude of the above classification accuracy is due to the large prediction in the nonrelapse class that as many as 269 people with nonrelapse status are appropriately predicted in the nonrelapse status class. Conclusion: There are four important variables included in the best MARS model that is age of first use of drugs, how to use drugs, marital status and jobs. The accuracy level of drugs user rehabilitation classification in East Java using MARS method produces accuracy of 95,71% and misclassification of 4,29%.
Keywords: Multivariate adaptive regression spline, Classification accuracy, Drugs user.
|
|---|
