Optimal Harvest date Prediction by Integrating Past and Future Feature Variables

Agriculture, especially horticultural farming is one of the most important issues in the world. On the business side, rural area has a disadvantage in the competitiveness with the region near the huge market by the transportation cost. To increase competitiveness, the countryside has adopted greenhouse cultivation to maximize the winter crop yield when there is a low yield near the metropolitan area. In addition, most of the farming environments are set up and manipulated by farmers for themselves. Consequently, they need not only knowledge about farming but also Information and Communication Technology (ICT) ability to be able to understand information. However, it is inefficient for new farmers to acquire both ICT skills and agricultural knowledge, so recent smart farms use AI technology to analyze obtained data. As a first step, to provide expertise and support for new farmers, we propose a method to predict the optimal harvest date of eggplant combining past pattern and future feature variables. That is because estimating accurate harvest date is important for crops that have a short period of optimum size, weight or quality. The proposed model is composed of pattern analysis and solar radiation prediction models, and it forecasts the growth rate as a crop’s response. We evaluated several methods and the result shows that the proposed method is an efficient tool for predicting the optimal harvest date in IoT-enabled smart greenhouse. This method can contribute to providing specialized and useful information for inexperienced farmers. Moreover, farmers advantage business contract, as estimating early optimal and more accurate harvest date.