SSE-Based Evolutionary Algorithm for Hyper-parameter Optimization of LightGBM on Paddy Rice Yield Prediction Problem.

One of the purposes of smart agriculture is to predict the yield of paddy rice using agricultural data using machine learning. LightGBM, one of the machine learning algorithms, is applied to the yield prediction problem of paddy rice in this paper. Since LightGBM has a large number of hyperparameters, the hyperparameter optimization using the stochastic schemata exploiter (SSE) is used. From the results of comparison with Genetic Algorithm (GA), it is confirmed that SSE has a fast convergence speed. In addition, it is found that the higher the mutation rate of SSE, the more converged to the global optimal solution without falling into the local solution.