A Genetic Programming-Driven Data Fitting Method

杨利霞

A Genetic Programming-Driven

Data Fitting Method

& G0 Q( N) G- e, j2 t, U% X
6 f# a, n% C+ F) H0 MData fifitting is the process of constructing a curve, or a set of mathematical functions, that has
2 x" N2 T/ U: t; l% H9 r+ e: }- dthe best fifit to a series of data points. Different with constructing a fifitting model from same type of function,
such as the polynomial model, we notice that a hybrid fifitting model with multiple types of function may have
a better fifitting result. Moreover, this also shows better interpretability. However, a perfect smooth hybrid
0 U9 G% ^: o% a: Y& H1 e! }+ ]fifitting model depends on a reasonable combination of multiple functions and a set of effective parameters.
8 R. a8 |# M% J2 @7 ?2 w4 [That is a high-dimensional multi-objective optimization problem. This paper proposes a novel data fifitting
$ N+ Q1 l. U1 c! D" k$ C  Mmodel construction approach. In this approach, the model is expressed by an improved tree coding expression
+ [* W7 L& A3 y5 Tand constructed through an evolution search process driven by the genetic programming. In order to verify
/ b2 k% I" \1 z7 zthe validity of generated hybrid fifitting model, 6 prediction problems are chosen for experiment studies. The
6 `: Z$ C4 P* d" Qexperimental results show that the proposed method is superior to 7 typical methods in terms of the prediction
5 R+ Z2 C! z5 ?8 a2 G' T" }3 Jaccuracy and interpretability.

6 s0 t* r' l  E2 N9 U: I) ?; Z