PENERAPAN METODE RANDOM FOREST DALAM DRIVER ANALYSIS

  • Nariswari Karina Dewi FMIPA IPB
  • Utami Dyah Syafitri FMIPA IPB
  • Soni Yadi Mulyadi PT. Ipsos Indonesia

Abstract

Driver analysis is one approach to know which  the greatest expalanatory variables influence the response variable. This analysis is well known in marketing research. In this area, explanatatory variables (X) and response variable (Y) ussually are measured by ordinal data and the relationship between those variables is non linier. One of the approach to build model on that situation is random forest. Two important things in random forest are size of random forest and sample size of X. In this research, we worked with  simulation to know the size of random forest which give higher accuration and more stabil. The simulation showed that the best condition achieved when the size of random forest is 500 and the sample size of X is 4.   

 

Key words : driver analysis, random forest, variable importance.

Section
Articles