Symbolic regression is an important but challenging research topic in datamining. It can detect the underlying mathematical models. Genetic programming(GP) is one of the most popular methods for symbolic regression. However, itsconvergence speed might be too slow for large scale problems with a largenumber of variables. This drawback has become a bottleneck in practicalapplications. In this paper, a new non-evolutionary real-time algorithm forsymbolic regression, Elite Bases Regression (EBR), is proposed. EBR generates aset of candidate basis functions coded with parse-matrix in specific mappingrules. Meanwhile, a certain number of elite bases are preserved and updatediteratively according to the correlation coefficients with respect to thetarget model. The regression model is then spanned by the elite bases. Acomparative study between EBR and a recent proposed machine learning method forsymbolic regression, Fast Function eXtraction (FFX), are conducted. Numericalresults indicate that EBR can solve symbolic regression problems moreeffectively.
展开▼
