Classifying Arabic Farmers’ Complaints Based on Crops and Diseases Using Machine Learning Approaches

Abstract

In this paper, two models are proposed to automatically categorize the farmers’ complaints on various diseases that may affect crops. In the first model, a complaint which is expressed in Arabic free text is classified into its respective crop and then into a particular disease, but in the second model, the complaint is classified directly into diseases. A dataset of farmers’ complaints described in Arabic script consisting of complaints from five different crops raised by farmers of different parts of Egypt is created. Separate lexicons are created for crops and their diseases by considering all possible technical terms related to a crop and its diseases including their possible slang synonyms. Each preprocessed complaint is represented in the form of a binary vector using the vector space model (VSM) with the help of crop lexicon so that machine learning techniques can be applied. Experiments are conducted on the dataset by varying the percentage of training with multiple trials using SVM and KNN classifiers. It has been observed from the results that the proposed model is performing on par with the human expert and can be deployable for real-time operations.