DisplayTitle

Classification of Dry-Milled Maize Grit Yield Groups Using Quadratic Discriminant Analysis and Decision Tree Algorithm

¹Office of the Texas State Chemist, Texas Agricultural Experiment Station, College Station, TX 77841-3160. ²Corresponding author. Phone: 979-845-1121. Fax: 979-845-1389. E-mail: tjh@otsc.tamu.edu ³USDA-ARS, Grain Marketing and Production Research Center, Manhattan, KS 66502. Names are necessary to report factually on available data; however, the USDA does not guarantee the standard of a product, nor does the use of the name by the USDA imply any approval of the product to the exclusion of others that may also be suitable. ⁴Department of Food Science and Technology, University of Nebraska-Lincoln, Lincoln, NE 68583-0919. A contribution of the University of Nebraska Agricultural Research Division, supported in part by funds provided through the Hatch Act. Mention of a trade name, proprietary product, or company name is for presentation clarity and does not imply endorsement by the authors or the University of Nebraska. ⁵Department of Agronomy, Kansas State University, Manhattan, KS 66506.

A genetically and environmentally diverse collection of maize (Zea maize L.) samples was evaluated for physical properties and grit yield to help develop a standard set of criteria to identify grain best suited for dry-milling. Application of principal component analysis (PCA) reduced a set of approximately 500 samples collected from six states to 154 maize hybrids. Selected maize hybrids were placed into seven groups according to their dry-milled grit yields. Regression analysis explained only 50% of the variability in dry-milling grit yield. Patterns of differences in the physical properties for the seven grit yield groups implied that the seven yield groups could be placed into two or three groups. Using two pattern recognition techniques for improving classification accuracy, quadratic discriminant analysis and the classification and regression tree (CART) model, dry-milled grit yield groups were predicted. The estimated correct classification rates were 69–80% when the samples were divided into three yield groups and 81–90% when samples were divided into two yield groups. The results indicated the comparable success of both techniques and the superiority of the decision tree algorithm to quadratic discriminant analysis by offering higher accuracy and clearer classification rules in differentiating among dry-milled grit yield groups.