Smart packaging is one of innovations in the field of packaging that can monitor and provide information to producers and consumers about the quality of the packaged product. Nowadays, there is no exact model to relate both of the changing between the label and the product quality. Thus, it is required an association between discoloration smart packaging indicator with the packaged product quality change. Parameter identification process can generate a lot of parameters, we need a method to select the important parameter attributes therefore it used a ReliefF (Reliable Eliminated of Feature) method. The objectives of this research were to identify the association parameters and to select the important attribute and sufficiently describing the relationship and furthermore this research was to develop an association model of discoloration of smart packaging indicator with natural dyes and the quality changes. The data were obtained from a previous research and discretization to classify the data °Hue, ΔE, consumer preferences and the total colony to nominal value. Selection of attributes by using the ReliefF method showed that there were three attributes with the highest weight value (Wi). Furthermore, this research generated important parameters and sufficiently describing the relationship of discoloration and product quality changes.