标题: A Genetic Programming-Driven Data Fitting Method [打印本页] 作者: 杨利霞 时间: 2020-11-10 16:01 标题: A Genetic Programming-Driven Data Fitting Method
A Genetic Programming-Driven
: H: G) E* d* H( \% D# c
Data Fitting Method
( W' A. }2 E; E& ?
7 e) G3 F3 c3 f/ R6 F2 }; Q+ B
3 @0 I8 U6 @; @: S. G
& v4 v4 N6 G# I8 s+ {9 y$ B
Data fifitting is the process of constructing a curve, or a set of mathematical functions, that has0 r7 |* y8 M6 Y+ v0 U1 W. g6 [- q: h
the best fifit to a series of data points. Different with constructing a fifitting model from same type of function,0 p0 r5 _7 [ y2 K
such as the polynomial model, we notice that a hybrid fifitting model with multiple types of function may have , A8 _/ r: ?6 S& La better fifitting result. Moreover, this also shows better interpretability. However, a perfect smooth hybrid 6 _* w; A1 x. Y& d8 G) M" a/ X/ gfifitting model depends on a reasonable combination of multiple functions and a set of effective parameters. D# g4 g# W4 m3 E3 D1 T( G3 k
That is a high-dimensional multi-objective optimization problem. This paper proposes a novel data fifitting6 q) j6 F: ~/ V/ W
model construction approach. In this approach, the model is expressed by an improved tree coding expression2 A8 ?( x/ D, j5 t/ I. B- e b# {
and constructed through an evolution search process driven by the genetic programming. In order to verify 7 e% P' @2 }) `the validity of generated hybrid fifitting model, 6 prediction problems are chosen for experiment studies. The* X6 h8 ?- ?% j) x
experimental results show that the proposed method is superior to 7 typical methods in terms of the prediction + N0 U" Z" |3 R: R( Taccuracy and interpretability. " b) V5 `4 l) y, [6 U
8 T0 W7 h: y* B; z% W; l: l, \9 Q* R0 k1 X( q& I: A. A3 C
