A Genetic Programming-Driven Data Fitting Method

杨利霞

A Genetic Programming-Driven

Data Fitting Method

  ]& `+ Q& b4 D, H) ~: }0 N6 t

, K9 E: C, A$ R- H7 J
Data fifitting is the process of constructing a curve, or a set of mathematical functions, that has
$ d0 W% o) J' G: F/ ~! q5 J5 o/ J0 vthe best fifit to a series of data points. Different with constructing a fifitting model from same type of function,
. H7 Y, a' S0 j2 a" K% {* Vsuch 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
* q" a9 K+ S& Yfifitting model depends on a reasonable combination of multiple functions and a set of effective parameters.
+ I' h' ~  v! jThat is a high-dimensional multi-objective optimization problem. This paper proposes a novel data fifitting
, D2 Q+ J" S& p7 a1 i  qmodel construction approach. In this approach, the model is expressed by an improved tree coding expression
/ v" y9 k" w3 Rand constructed through an evolution search process driven by the genetic programming. In order to verify
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
accuracy and interpretability.

2 D4 c8 D- k* B2 f) r' h9 y
8 E( O; y) [- N/ ?+ D