Application of data mining and adaptive neuro-fuzzy structure to predict color parameters of walnuts (Juglans regia L. )

Quality is the primary factor designating consumer satisfaction and the market price of agricultural commodities. Color and general appearance are the basic quality indicators for agricultural products. Surface colors are assessed through colorimetric measurements including 1*, a*, and b* color parameters. In the present study, 1*, a*, and b* color parameters of Bilecik, Fernette, Fernor, Kaman-1, Maras-12, Maras-18, Sunland, Sen-2, Yalova-1, and Yalova-3 walnut cultivars (color parameters of 100 randomly selected walnuts from each cultivar) were measured with a chroma meter (CR-5 Konica Minolta). Based on L*, a*, and b* measurements, equations from which color index (CI), chroma (C*), and hue (h*) angle parameters could be calculated were developed with the Find Laws algorithm of PolyAnalyst. The color parameters obtained from these newly developed equations were used in training of adaptive neuro-fuzzy structure. Then color index (Cl), chroma (C*), and hue (h*) angle parameters were predicted by adaptive neuro-fuzzy approach. Root mean square error values of the adaptive neuro-fuzzy-based approach were respectively identified as 0.02 for Bilecik, 0.01 for Fernette, 0.02 for Fervor, 0.01 for Kaman-1, 0.01 for Maras-12, 0.01 for Maras-18, 0.01 for Sunland, 0.01 for Sen-2, 0.01 for Yalova-1, and 0.01 for Yalova-3 walnuts. The obtained equations can be used as a viable alternative instead of equations that vary depending on whether a* and b* are negative or positive.