Nutrient recommendation model for plantain (Musa AAB) cultivation based on the analysis of soil variables in the Department of Risaralda, Colombia

Abstract In Colombia, one of the foods that guarantees food security is the plantain plant. Given its nutritional properties and low cost, it is grown mainly by peasant families with low financial resources. One of the greatest problems in the production chain of this crop is the incorrect selection of soils and integrating the field with the technology. This research proposes a recommendation model for the efficient choice of soil for this crop. This model analyzes a set of data (pH, organic material, phosphorus, sulfur, potassium, calcium and magnesium) product of the soil analysis in the laboratory and generates a list of recommended soil values adjusted for this crop and their respective correlation coefficient (similarity) with the soil values of the land requested by the farmer. For this, a similarity matrix was constructed by applying Pearson’s coefficient. Additionally, the model allows us to obtain the probability of compatibility between the identified edaphic values and the values needed by the soil for this crop. The KNN, SVM, C5.0 and logistic regression algorithms were compared using the following metrics: AUC, accuracy and error rate, F1 index, recall and precision. The KNN model. presented the best indicators. Finally, the recommendation model was built using the R programming language and implemented through a software application. This research aims to help farmers choose the right soil for cultivation by suggesting nutrients that should be applied to increase soil quality and adopt strategies that reduce soil nutrient loss.