Plant Growth Prediction Method for Plant Factories Using LSTM Algorithm

Abstract

In recent years, the ratio of new type agriculture has been increased. Agriculture in plant factories has been attracting attention. However, these types of agriculture are not as profitable as conventional ones. Therefore, it is necessary to control the cultivated schedule as one of the ways to improve the profitability. In this paper, we present the method for plant growth prediction in plant factories. Microcomputers and sensors are used to measure the data of the cultivation environment, and we predict size and weight for vegetable by using LSTM algorithm. From experimental results epochs of 20, 50, and 70, the best accuracy is obtained at epoch number of 50. Then MSE and MAE are 0.077348 and 0.187984, respectively. The coefficient of determination is as low as −0.529. MSE and MAE are 0.165420 and 0.328250, respectively, which were worse than size, and the coefficient of determination exceeded −2. From about results, the prediction of the size is mostly completed. In the future the prediction accuracy of the weight needs to improve.