In this paper, a method based on the Genetic Algorithm (GA) and SVD-QR method is proposed to construct an appropriate fuzzy system for data classification. In this method, an individual of the population in the GA is used to determine a fuzzy partition such that some rough fuzzy sets of each input variable are obtained. In order to extract significant fuzzy rules from the rule base of the defined fuzzy system, the SVD-QR method is applied to remove unnecessary fuzzy rules such that the constructed fuzzy system has a low number of fuzzy rules. A fitness function in the GA is considered to guide the search procedure to select an appropriate fuzzy system such that the number of correctly classified patterns are maximized and the number of fuzzy rules is minimized. Finally, a classification problem is considered to illustrate the effectiveness of the proposed method.