This paper deals with multi-attribute classification problem based on dynamical multi-objective optimization approaches. The matching of attribute is seen as objective of the problem and user preferences are uncertain and changeable. Traditional sum weighted method and simple evolutionary algorithm are employed for experimental study over practical industry product classification problems. A integrate system framework is proposed to realize the dynamical model for multi-objective optimization. The experimental results show that classification performance system can be improved under the dynamical system framework according to user preference.