Articles
Prediction of tomato quality traits using machine learning models
Article number
1445_36
Pages
257 – 264
Language
English
Abstract
Tomatoes are a globally important resource in food processing, with significant implications for nutrition, agriculture, and health.
Understanding tomato ripening is essential, as it affects vital qualities like Brix, lycopene, and fruit colour.
Amidst global population growth and climate challenges, machine learning (ML) models emerge as promising tools for future in predicting tomato quality.
This study utilized a data set comprising physicochemical traits and environmental factors for 48 tomato cultivars grown across 20 locations over five seasons in Hungary.
We trained two machine learning models, extreme gradient boosting (XGBoost) and Artificial Neural Networks (ANNs), to predict three key tomato quality metrics: °Brix, lycopene content, and the chromaticity ratio (a/b ratio). The results demonstrated that in predicting both °Brix and lycopene levels, XGBoost was more effective than the ANN model.
In case of °Brix, XGBoost recorded a high R2 value of 0.98 and a minimal RMSE of 0.07, exceeding the performance of ANNs which attained an R2 of 0.89 and a RMSE of 0.17. Similarly, in the prediction of lycopene content, XGBoost achieved an R2 of 0.87 and an RMSE of 0.61, bettering the ANN’s R2 of 0.84 and RMSE of 0.86. Additionally, XGBoost excelled in predicting the a/b ratio with an R2 of 0.93 and a minimal RMSE of 0.03, while the ANN model displayed limitations, resulting in a negative R2 value of -0.35.
Understanding tomato ripening is essential, as it affects vital qualities like Brix, lycopene, and fruit colour.
Amidst global population growth and climate challenges, machine learning (ML) models emerge as promising tools for future in predicting tomato quality.
This study utilized a data set comprising physicochemical traits and environmental factors for 48 tomato cultivars grown across 20 locations over five seasons in Hungary.
We trained two machine learning models, extreme gradient boosting (XGBoost) and Artificial Neural Networks (ANNs), to predict three key tomato quality metrics: °Brix, lycopene content, and the chromaticity ratio (a/b ratio). The results demonstrated that in predicting both °Brix and lycopene levels, XGBoost was more effective than the ANN model.
In case of °Brix, XGBoost recorded a high R2 value of 0.98 and a minimal RMSE of 0.07, exceeding the performance of ANNs which attained an R2 of 0.89 and a RMSE of 0.17. Similarly, in the prediction of lycopene content, XGBoost achieved an R2 of 0.87 and an RMSE of 0.61, bettering the ANN’s R2 of 0.84 and RMSE of 0.86. Additionally, XGBoost excelled in predicting the a/b ratio with an R2 of 0.93 and a minimal RMSE of 0.03, while the ANN model displayed limitations, resulting in a negative R2 value of -0.35.
Authors
O. Mhamdi, S. Takács, G. Palotás, R. Ilahy, L. Helyes, Z. Pék
Keywords
extreme gradient boosting, artificial neural network, lycopene, quality forecast
Online Articles (49)