Comparative Analysis for Classification of Rice Genotype Using Statistical and Artificial Neural Network Models on the Basis of Maturity

Abstract

Classification is that arena of science which deals with grouping of objects on the basis of information available about objects. It plays a significant role for planning purposes in agriculture system. This study aims at comparing classification techniques through statistical as well as artificial neural network models for the primary data related to 140 rice genotypes from the trial laid in SKUAST, Jammu on the basis of maturity. And, the characters like yield per plant, number of days for 50% flowering, number of days for full flowering, plant height, number of effective tillers per plant, panicle length, grain length, grain width and ratio of grain length & grain width acts as supporting variables for classification. The statistical model used for classification technique was ordinal logistic regression model whereas in case of artificial neural network, multilayered perceptron neural network was used. The class variable number of days for 50% flowering was categorized into three classes and was considered as dependent variable and all other characters as independent variables. The ability measures of classification such as Accuracy Rate, Kappa Statistics, Average precision and Average Recall were used for testing samples. It is observed that Multilayered Perceptron Neural Network performed better than Ordinal Logistic Regression for classifying rice genotypes for different classes of number of days for 50% flowering on the basis of classification ability measures. Number of days for full flowering was found to be important attributing character for classification on the basis of maturity.