A Machine Learning Model for Early Prediction of Crop Yield, Nested in a Web Application in the Cloud: A Case Study in an Olive Grove in Southern Spain

Abstract

Predictive systems are a crucial tool in management and decision-making in any productive sector. In the case of agriculture, it is especially interesting to have advance information on the profitability of a farm. In this sense, depending on the time of the year when this information is available, important decisions can be made that affect the economic balance of the farm. The aim of this study is to develop an effective model for predicting crop yields in advance that is accessible and easy to use by the farmer or farm manager from a web-based application. In this case, an olive orchard in the Andalusia region of southern Spain was used. The model was estimated using spatio-temporal training data, such as yield data from eight consecutive years, and more than twenty meteorological parameters data, automatically charged from public web services, belonging to a weather station located near the sample farm. The workflow requires selecting the parameters that influence the crop prediction and discarding those that introduce noise into the model. The main contribution of this research is the early prediction of crop yield with absolute errors better than 20%, which is crucial for making decisions on tillage investments and crop marketing.