Application of Artificial Neural Network (ANN) and Logistic Regression for Predicting Weeds Presence in Dryland Chickpea Fields of Kurdistan Province

A survey was conducted to compare the potential of ANN and logistic regression in predicting weed presence of 33 dryland chickpea fields in Kurdistan province. Climatic and edaphic factors as independent variable and presence or absence of weeds with highest abundance as dependent variables were entered in the logistic regression and ANN models. The developed ANN was a Multi Layer Perceptron with nine neurons in the input layer, one and two hidden layer(s) of various numbers of neurons and two neurons in the output layers. Catchweed (Galium aparine L.) and field bindweed (Convolvulus arvensis L.) with the highest abundance indices were the dominant weeds in the chickpea fields. The logistic regression did not fit a model for catchweed, however, the ANN could develop the best suited models for predicting the catchweed and field bindweed presence in dryland chickpea fields. Sensitivity analysis revealed that altitude and rainfall are the most significant parameters in modeling weed presence in dryland chickpea fields. For the optimal model, the values of the model’s outputs correlated well with actual outputs and its application for this purpose is recommended.