Bayesian Ridge Algorithm for Brix Prediction in Industrial Tomato

Abstract

Tomato is one of the most significant vegetables in the world. Specifically, for the industrial tomato cultivation, the product is harvested when °Brix are at their peak. Technological advancements nowadays have made Decision Support Systems, based on Machine Learning Algorithms more applicable in a daily basis. Sustainable agriculture is evolving since farmers could be advised by this technology in order to take the best decision for their crops. Farmers who adopt this kind of technology will be able to know the quality of tomatoes. The implementation of a Decision Support System capable to predict the °Brix was conducted, based on various data from previous years, such as quality characteristics, the tomato hybrid used, weather conditions and soil data from the selected fields. Data came from fields from 6 different regions in Peloponnese, Greece over 3 cultivation periods. 12 different algorithms were tested in order to find which is the best one in terms of efficiency. Results of this research showed that the predicted °Brix were following the same pattern as the actual °Brix. This means that the DSS could advise the farmer about the ideal harvesting period where the °Brix will be maximized. The use of this DSS using real time weather data as an input will be a valuable tool for the farmers.