Developing a mobile application to predict maize storage losses
Abstract
Maize is among the major cereals produced in the world and yet suffers the highest percentage of post-harvest loss compared to other cereals with an annual estimate of 17% in Uganda. These losses are cumulative across the different stages of post-harvest handling but most of the losses occur during storage. Efforts to reduce these losses have not reaped considerable success owing to the inability of farmers and other stakeholders in the post-harvest chain to select the best storage method and optimize storage conditions. Limited interventions have been done to optimize storage conditions and predict possible maize storage losses. The main aim of this project was to develop a mobile application to predict maize storage losses. The research objectives were to develop, evaluate, and deploy a maize storage loss predicting mobile application. The predictive model was developed using machine learning algorithm and evaluated using the evaluation metrics of mean squared error, R2 value, and adjusted R2 value. The developed model had a root mean squared error was 10.35kg, R2 value of 0.784, mean absolute error of 8.256 and root relative square error of 15.25. The developed model and mobile application can be used by different stakeholders such as food retailers, institutions that store food in bulk like the military, schools, and food relief organizations. The application can be used to optimize storage conditions to reduce grain loss in storage.