Smart Agriculture using Ensemble Machine Learning Techniques in IoT Environment
Liyakathunisa Syed
Abstract
Crop yield is a serious concern for farmers due to irregular irrigation, soil erosion, uncontrolled seed planting, severe weather conditions, peasants, and unpredicted locust infestations. The lack of current data and the intricacy of traditional agriculture leads to inefficiency and excessive operation costs. Smart farming with Artificial Intelligence techniques and the Internet of Things (IoT) in agriculture overcome these challenges by connecting all the accessible data sources into a single, fully effective functional unit. Smart agriculture helps in the sustainability of food production by using minimum resources such as water, fertilizer, and seeds. It provides a better understanding of the soil, crop, and changing weather. Furthermore, the sensors in the system assist in monitoring and controlling the resources. In this research, we present ensemble-based machine learning approaches in the IoT environment to predict crop yields and enable sustainable farming by guiding the farmers to grow the correct crop during the correct season to increase their yields. The proposed method uses a novel ensemble-based machine learning classification approach with two levels of predictions-level-0 prediction includes (Logistic Regression (LR), Classification and Regression Trees (CART), Support Vector Machines (SVM), and K-Nearest Neighbors (KNN)) classifiers are input as features to the level-1 meta classifier (Random Forest) to detect the distinct categories of different crops. Its predictions can be extremely close to the ground truth, which accurately predicts 22 categories of crops with 99% accuracy. Hence, the proposed Smart agriculture system can assist farmers in improved yield production and remote monitoring at low cost.