A Genetic Programming-Driven Data Fitting Method

杨利霞

A Genetic Programming-Driven

Data Fitting Method

; G3 p' X& a' N0 ~& N' {
7 `, \1 k. y2 N" `/ O, ~1 k5 C

! F- o( V' A1 E- hData fifitting is the process of constructing a curve, or a set of mathematical functions, that has
0 y. s7 H  B; o) |5 Xthe best fifit to a series of data points. Different with constructing a fifitting model from same type of function,
. c! k8 E& }: P1 ^! N/ ?5 h) lsuch 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
fifitting model depends on a reasonable combination of multiple functions and a set of effective parameters.
  Y$ I3 U8 E8 _7 T& C" O' FThat is a high-dimensional multi-objective optimization problem. This paper proposes a novel data fifitting
: h* T: c  E9 W' X, Y7 m3 Pmodel construction approach. In this approach, the model is expressed by an improved tree coding expression
, R7 E3 }  h2 |9 |$ b( i0 @and constructed through an evolution search process driven by the genetic programming. In order to verify
" C( o2 `, ?! p% V1 X5 ?the validity of generated hybrid fifitting model, 6 prediction problems are chosen for experiment studies. The
experimental results show that the proposed method is superior to 7 typical methods in terms of the prediction
1 k7 n/ W* P5 H0 a+ H* p2 Y; eaccuracy and interpretability.
7 x3 n1 `0 N( y; A0 l$ m

6 q* r( B8 M& l( ?# [


# e, H) W" u% z2 `; m