Sensitive information identification method combining MTL and fuzzy classification

To address the limitations of traditional sensitive information classification methods when handling complex data structures and fuzzy boundaries, a method combining multi-task learning and fuzzy logic is proposed. The method constructs a multi-task(MTL) model based on LSTM and attention mechanisms, with classification adjustments made through fuzzy logic. Comparative experimental results show that the proposed method significantly outperforms traditional methods in metrics such as accuracy, precision, recall, F1 score, and AUC, achieving an accuracy of 99.97%, an F1 score of 0.9993, and an AUC of 0.9994. Compared to models like Logistic Regression, Random Forest, and Neural Networks, this method demonstrates a clear performance advantage. Ablation experiments further validate the contribution of multi-task learning and fuzzy logic in improving model performance. Overall, the model performs excellently across various evaluation metrics, proving its effectiveness and practicality in sensitive information recognition tasks.