Development of predictive model based mobile application for maturity stage identification of Indian traditional red bananas

This research work proposes to develop a mobile application for predicting the maturity stage of Indian traditional var. red banana fruits based on user-provided parameters including fruit weight, length, caliper, and fruit girth. A customized data set was prepared by measuring the various parameters at regular intervals with different maturity stage (0-100%). This data set comprising of measurements for all different maturity stages was utilized to develop and evaluate various regression models, including linear, polynomial, ridge, gradient boosting, random forest, and decision tree regression, to identify the most effective model for maturity stage prediction. The results highlighted that Ridge regression as the optimal choice, exhibiting a high R2 value of 0.99764 and minimal Mean Squared Error (MSE) of 1.92, indicating its superior handling of multi collinearity and predictive accuracy. Random forest regression also performed equivalently well with lesser R2 value of 0.9884. However, the higher MSE (9.4058) of random forest regression model suggested that ridge regression model is the best suitable for maturity stage identification of traditional Indian red banana. A mobile application was developed with the ridge regression based predictive model using flutter which had highest accuracy during testing and validation. Deployment of the Ridge regression-based predictive model into the mobile application offers a practical tool for farmers and stakeholders to predict the maturity stage accurately, aiding decision-making during cultivation, harvest, post harvest handling and marketing.