Analysis of Decision Tree and Random Forest Algorithms for Nutrient Deficiency Classification in Citrus Leaves through Image Processing

The research presents a hybrid approach to identify and categorise nutritional deficiency syndrome in citrus leaves using image processing and machine learning. The method includes processing images, segmenting images and extracting color-texture information. The Decision Tree and the Random Forest algorithms use various statistical measures for classification, including mean, skewness, variance, entropy, maximum probability, angular second moment, contrast, correlation, dissimilarity, energy, and homogeneity. The test results show a range of performance levels, with an accuracy of 80.28% being attained by utilising an 80:20 data ratio in the Decision Tree approach. The Random Forest method’s highest accuracy, 91.25%, was reached using an 80:20 ratio. In conclusion, the accuracy of the Random Forest approach regularly beats that of the Decision Tree method.