A cost-sensitive bayesian tobacco grade identification system under unbalanced data
By constructing a cost-sensitive Bayesian tobacco grade identification system under imbalanced data, the problems of low identification rate of high-value tobacco leaves, high economic cost of misclassification, and lack of model interpretability in tobacco grade classification are solved. It achieves accurate identification of high-value tobacco leaves and minimizes economic costs, and provides transparency and traceability in decision-making.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- KUNMING UNIV OF SCI & TECH
- Filing Date
- 2026-04-28
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for grading tobacco leaves suffer from problems such as low recognition rate of high-value tobacco leaves, high economic cost of misclassification, lack of model interpretability, and poor adaptability of Bayesian networks. In particular, it is difficult to achieve accurate recognition and minimize economic costs under imbalanced data.
By employing multi-level integrated resampling, cost-sensitive Bayesian network learning, and probabilistic inference techniques, a cost-sensitive Bayesian tobacco grade identification system under imbalanced data is constructed. This system includes data preprocessing, multi-level integrated resampling, cost-sensitive Bayesian network learning, inference prediction, and evaluation modules, enabling accurate identification of high-value tobacco leaves and transparent and traceable decision-making.
It improves the identification accuracy and recall rate of high-value tobacco leaves, reduces the economic losses from misclassification, provides interpretability and traceability of the model, adapts to the asymmetric cost of misclassification, and significantly outperforms the traditional model.
Smart Images
Figure CN122413136A_ABST