Analysis of TCM Data Based on Partial Least Squares within Random Forest

Abstract

Abstract. Partial Least Square (PLS) seems hard to adapt to the characteristics of the nonlinear data due to its own linear feature. However, Random Forest Algorithm(RFA), which is assembled by several classifiers, is adaptive and suitable to nonlinear data. Based on this, a new method fusing RF into PLS is proposed, which build Random Forest through the principal components and the dependent variable extracted from PLS, and use the residual information to build Random Forest recursively until accuracy conditions are met. Using the data of the maxingshigan decoction of the monarch drug to treat the asthma or cough and some datasets in the UCI Machine Learning Repository, the results show that the improved algorithm has a certain degree of correctness and validity.