Yield Estimation and Clustering of Chickpea Genotypes Using Soft Computing Techniques

Crop growth is a multifactorial nonlinear process and different kinds of models have been developed to predict crop yield. In recent years, crop growth models have become increasingly important as major components of agriculture-related decision-support systems. Moreover, clustering is a multivariate analysis technique widely adopted in agricultural studies. Using this method, different genotypes (accessions) of crops can be classified and characterized. This paper discusses the use of soft computing techniques such as artificial neural networks (ANN) and fuzzy logic based approaches in regression and clustering problems. The ANN technology was used for modeling the correlation between crop yield and 10 yield components of chickpea (Cicer arietinum L.). Also, the fuzzy c-means (FCM) clustering technique was used for the classification of 362 chickpea genotypes based on their agronomic and morphological traits. The ANN performed very well. Among the various ANN structures, a model of good performance was produced by 10-14-3-1 structure with a training algorithm of back-propagation and hyperbolic tangent transfer function in the hidden and output layers. The model was able to predict the chickpea yield data of 0.32 to 14.38 g plant-1 with a RMSE and T value of 0.0195 g plant-1 and 0.988, respectively. T statistics measures the scattering around line (1:1). When T is close to 1.0, the fitting is desirable. The mean absolute error, relative error, and coefficient of determination between actual and predicted data were 0.0109 g plant-1, -1.07%, and 0.991, respectively. The ANN model predicted 90.3% of the yield data with relative errors ranging between ±5%. The consequent reduction in the number of training data from 250 to 50, decreased the training RMSE, but increased the prediction error. It was found that even with a few number of patterns in the training dataset (50 patterns), the prediction error of the ANN model was in the range of acceptance for yield modeling. Obviously, with 25 x 103 iterations, the ANN models with 5 and 10 input variables gave almost the same estimation of the chickpea yield. The results of clustering showed that the FCM clustering technique can be successfully applied to classify chickpea genotypes in terms of agronomic and morphological traits.