Related Experiment Video
Updated: Jan 7, 2026
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
5.2K
A Precise Apple Quality Prediction Model Integrating Driving Factor Screening and BP Neural Network.
Junkai Zeng1,2, Mingyang Yu1,2, Yan Chen1,2
1College of Horticulture and Forestry Science, Tarim University, Alar 843300, China.
Plants (Basel, Switzerland)
|December 31, 2025
Summary
This study develops a Back Propagation (BP) neural network model to predict apple quality, using photosynthetic and nitrogen metabolism data. The model accurately predicts Vitamin C, Soluble Saccharides, and Titratable Acid, enhancing smart orchard management.
Area of Science:
- Horticulture
- Plant Physiology
- Agricultural Engineering
Background:
- Apple fruit quality is crucial, determined by Vitamin C (VC), Soluble Saccharides (SSs), Titratable Acid (TA), and their ratio (SSs/TA).
- Accurate prediction of these quality parameters is essential for effective orchard management and consumer satisfaction.
Purpose of the Study:
- To establish a precise prediction model for apple fruit quality indicators using Back Propagation (BP) neural networks.
- To analyze the intrinsic relationships between fruit quality and the physiological characteristics of apple trees, specifically photosynthetic and carbon-water-nitrogen metabolism indicators.
- To develop a simplified and efficient prediction model through feature screening for improved accuracy in quality prediction.
Main Methods:
- Systematic measurement of fruit quality indicators, leaf photosynthetic parameters (Pn, Gs), canopy structure, and carbon-water-nitrogen metabolism indicators in 'Fuji' apples.
- Correlation analysis to identify key influencing factors on fruit quality.
- Construction and optimization of BP neural network models, including feature importance analysis, sensitivity analysis, and ablation experiments for feature subset selection.
Main Results:
- Photosynthetic parameters (Pn, Gs, SUE, DUE) and nitrogen metabolism (N, NLT) were significantly correlated with key apple quality indicators (VC, SSs, TA, SSs/TA).
- The optimal BP neural network model achieved high predictive accuracy for VC, SSs, TA, and SSs/TA, with validation set R² values ranging from 0.86 to 0.89.
- A simplified model developed through feature screening further improved prediction accuracy for VC (R² = 0.93) and reduced errors (MAE, MAPE).
Conclusions:
- Photosynthetic characteristics and nitrogen metabolism are fundamental physiological drivers of apple fruit quality.
- The BP neural network model provides a highly accurate and effective tool for predicting apple quality.
- Feature refinement significantly enhances the predictive performance of the BP model, offering a valuable approach for smart orchard management and precise quality regulation.
