.png){kind=icon}
.png){kind=icon}
Return
Predicting, modelling, and optimising the properties of edible films using machine learning algorithms
.png){kind=icon}
PRE
AI
G
H
DOI：10.1007/s11694-026-04825-6.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Determining the best formulation for edible film production through traditional trial-and-error methods is a labour-intensive, time-consuming, and costly process. To overcome these challenges, an innovative machine learning-based approach was adopted in this study. Within the scope of the research, the physical, mechanical, and chemical properties of film samples (F0, F1, F5, F10) produced in a laboratory environment using linseed gum and onion peel waste were comprehensively characterised. Subsequently, various machine learning algorithms, including Random Forest, SVR, Linear Regression, Bayesian Regression, and k-NN, were applied to this complex multivariate dataset to accurately predict critical performance characteristics of the film, such as elongation at break, light transmittance, antioxidant activity, and total phenolic content. Among the applied algorithms, Random Forest achieved the highest performance, with an R² score of 0.8655. To improve the interpretability of model performance, SHAP and LIME methods were applied to analyse the quantitative effects of input variables on output variables. SHAP results revealed that the amount of onion peel extract was the dominant determinant of antioxidant capacity and phenolic content. At the same time, LIME analyses showed how certain variables influenced the prediction in individuals close to the optimal solution. For the explainable analysis of film properties, Accumulated Local Effects (ALE) and Permutation Feature Importance (PFI) methods were applied. PFI analysis quantitatively revealed the relative effects of formulation parameters on film performance, while ALE analysis showed the direction of the parameters and their nonlinear behaviour within specific ranges. The findings revealed that the amount of onion peel extract was decisive for total phenolic content and antioxidant capacity.
Keywords:
Data-driven optimisation
Edible film
Machine learning
Sustainability
Journal
J
IF:
3.3
Papers:
1.3K
Citations:
1.1W
Organization
Cited Papers