Sprinkler Irrigation Automation System to Reduce the Frost Impact Using Machine Learning





Frost, machine learning, sprinkler irrigation, random forests, linear regression, decision trees


Frosts reduce the ambient temperature to the freezing point of water, affecting the agricultural sector and the integrity of plant tissues, severely damaged by freezing, destroying plant cells. In addition, losses are generated in the economy due to the death of cattle due to cold, hunger, diseases, etc. Latin America is a region that depends, to a considerable extent, on its crops for its consumption and export, so frost represents an urgent problem to solve, considering that in Perú the area of agriculture is not technical. Among the methods most used by farmers is anticipated irrigation, through automatic learning techniques, which allows predicting the behavior of a variable based on previous historical data. In this paper, sprinkler irrigation is implemented in crops exposed to frost, using an automated system with machine learning techniques and prediction models. Therefore, three types of models are evaluated (linear regression, random forests, and decision trees) to predict the occurrence of frosts, reducing damage to plants. The results show that the protection activation indicator from 1.1°C to 1.7°C was updated to decrease the number of false positives. On the three models evaluated, it is determined that the most accurate method is the Random Forest Regression method, which has 80.91% reliability, absolute mean error, and mean square error close to zero.






Original Scientific Papers