Data Mining is the process of extracting useful hidden knowledge from large volumes of data and its results can be used in decision support systems. Several data mining algorithms have been developed; one example is the association rule mining, which discovers associations among items encoded within a database. When the values of the attributes in a database are continuous such as height, length or weight, their domain is usually discretized into several intervals, as a result, such attributes are handled as discrete attributes. In this study, a new approach of mining association rules for handling continuous attributes that does not use any discretization is proposed. The methodology is based on a new graph-based evolutionary algorithm named "Genetic Network Programming (GNP)" and fuzzy membership functions. Our data mining method first needs to transform the continuous values in transactions into linguistic terms, then judge them and find association rules using GNP. GNP represent its individuals using graph structures that evolve in order to find a solution; this feature contributes to creating quite compact programs and implicitly memorizing past action sequences. The proposed method can measure the significance of the extracted association rules by using support, confidence and χ2 test, and obtains a sufficient number of important association rules in a short time.