Dự đoán giá cổ phiếu trên thị trường chứng khoán Việt Nam bằng phương pháp lai GA-SVR

Abstract

Stock  price  prediction  is  an  interesting problem  that  has  attracted  much  attention  from  bothinvestors  and  researchers.  There  are,  however,  not many researchs in this field with Vietnam stock market because  this  market  is  still  nascent  and  high  nonstationary.  In  this  paper,  we  propose  a  hybrid approach,  which  integrates  Genetic  Algorithm  (GA) with  Support  Vector  Regression  (SVR)  to  predict Vietnam stock price. In this approach, GA solves two problems  simultaneously:  finding  SVR’s  optimal parameters and feature selection. Then, SVR’s optimal parameters  and  selected  features  serve  as  input  fortraining  SVR  model.  Our  experimental  results  show that  the  hybrid  GA-SVR  approach  outperforms  SVR, Artificial  Neural  Network  (ANN)  and  can  be  used  in practice to gain profit.