Approach of Analysis of Data Mining Prediction In Earthquake Case Using Non Parametric Adaptive Regression Method

Abstract

Data Mining is the process of finding certain patterns and knowledge from big data. In general, the data mining process can be grouped into two categories, namely descriptive data mining and data mining prediction. There are several Math functions that can be used in the data mining process, one of which is the Classification and Regression function. Regression Analysis is also called Prediction analysis, which is a statistical method that is widely used to investigate and model relationships between variables. Regression analysis to estimate the regression curve can be done by analyzing Nonparametric Regression. One well-known method in non-parametric regression is MARS (Multivariate Adaptive Regression Spline). The MARS method is used to overcome the weaknesses of the Linear Regression method. The use of a stepwise backward algorithm with the CQP quadratic programming framework (CQP) from MARS resulted in a new method called CMARS (Conic Multivariate Adaptive Regression Splines). The CMARS method is able to model high dimensional data with nonlinear structures. The flexible nature of the CMARS model can be used in the process of analyzing earthquake predictions, especially in Lombok, West Nusa Tenggara. Test results Obtained a mathematical model of four independent variables gives significant results to the dependent variable, namely Peak Ground Acceleration (PGA). Contributions of independent variables are the distance of the epicenter 100%, magnitude 31.1%, the temperature of the incident location 5.5% and a depth of 3.5%.