Stacking Approach for Crop Recommendation Using Meta Model

Agriculture is the most important aspect of nation development. If the citizens of a nation have ample food then it not only helps them to work efficiently but also contributes to the growth of farmers. In developing nations like India, the demand for crops is rising. As countries like India and China are densely populated, the need for food is high and it increases day by day. The need for crop recommendation is due to less availability of land for farming. If the land for agriculture is limited then it is very essential to identify whether the selected crop is suitable for farming. Supervised machine learning algorithms are used for crop recommendation. Generally, classification techniques like Support Vector Classification (SVC), Random Forest ensemble algorithm, and Decision Trees Classifiers are used for crop recommendation. In this hybrid approach, we have used a stacking algorithm. The stacking algorithm combines the predictions of various machine learning algorithms. In our stacking algorithm, we have considered the meta-model and base model. The output of the first base model acts as the input for its respective meta-model. We have used the artificial neural network as the metamodel. We have proposed three stacking algorithms. In the first stacking algorithm, we have used Random Forest and Decision Tree Classification as the base model whereas the artificial neural network is the meta model. For the second model, we have combined Support Vector Machine (SVM) and Random Forest Classification as the base model and the artificial neural network as the meta-model. The third algorithm we proposed is the combination of a Decision Tree Classifier and SVM with a Decision Tree Classifier working as a meta-model. We obtained metrics like precision, accuracy, recall, and f-measure. The second stacking model gives the best results for the crop recommendation with 99.34% accuracy.