Crop Recommendation via Clustering Center Optimized Algorithm for Imbalanced Soil Data
Aoqi Liu, Tao Lü, Bufan Wang, Chong Chen · 2020
Making regional crop recommendation based on the content of soil elements and nutrient cycle periods is an efficient and environmental-friendly way in precision agriculture. However, real-world soil data distribution is always imbalanced, which seriously affects the performance of the learning-based prediction model. In order to solve this problem, this paper proposes a clustering center optimized algorithm by Synthetic Minority Over-sampling Technique (SMOTE). Firstly, the algorithm analyzes the original sample points and selects the density-based clustering centers. Then, it uses clustering center to generate the minority samples to ameliorate imbalanced data distribution. Finally, the ensemble algorithm is used to train the prediction model for accurate prediction. Experimental results show that the prediction accuracy of the proposed algorithm outperforms other state-of-the-art prediction models over imbalanced soil data.