标题: A Genetic Programming-Driven Data Fitting Method [打印本页] 作者: 杨利霞 时间: 2020-11-12 16:33 标题: A Genetic Programming-Driven Data Fitting Method
A Genetic Programming-Driven
3 _4 c( ?7 r8 D/ {
Data Fitting Method
. e: Q. Y/ I5 Z" L" {9 H . A5 s9 Y. s& u9 S. J; k9 G; e) K( N6 D5 T* Q! x( k% t9 }
, A& \; Z8 O0 n/ \
Data fifitting is the process of constructing a curve, or a set of mathematical functions, that has4 ?! O3 X7 |$ D6 d
the best fifit to a series of data points. Different with constructing a fifitting model from same type of function,$ F5 f2 u* c* m
such as the polynomial model, we notice that a hybrid fifitting model with multiple types of function may have : O2 {) Z6 r% @- u& u: r9 {! aa better fifitting result. Moreover, this also shows better interpretability. However, a perfect smooth hybrid , [- t6 y* |: |" G0 ]4 dfifitting model depends on a reasonable combination of multiple functions and a set of effective parameters.5 r' |* O& w7 y# z( M
That is a high-dimensional multi-objective optimization problem. This paper proposes a novel data fifitting& e$ }# d5 T( \) X$ Y3 C' j
model construction approach. In this approach, the model is expressed by an improved tree coding expression 2 z4 B- _ ^" A" dand constructed through an evolution search process driven by the genetic programming. In order to verify ]% Q" I0 d* P, R7 S
the validity of generated hybrid fifitting model, 6 prediction problems are chosen for experiment studies. The( J2 b9 _% t; t k M8 {* \
experimental results show that the proposed method is superior to 7 typical methods in terms of the prediction& Q' C% S! v! y! _2 I/ Z9 |
accuracy and interpretability5 Z* \. R" s* c. }: Z/ ~8 e
$ l" _7 Y5 a9 y 3 ~$ l, y+ Q% V) V* b' p) T1 N- n & j4 q/ I5 |8 B3 M9 g/ r+ J. L 1 D( j j: j p1 N1 S" K 8 c3 y4 D7 V' l: c5 r, E
