Abstract: Partial Least- Squared(PLS) Regression can resolve multicollinearity of importance. As a newtype of searching method for global optimization, genetic algorithm has some merits such as intelligence searching, parallel mode and robust. After modeling by PLS regression, the coefficient is re- evaluated by the genetic algorithms depended on its effective self- adapting global searching optimization, at last PLS regression model based on genetic algorithms is obtained. The modeling is used and the results show that the modeling attains good simulating effect and forecasting precision.