Machine Learning and Cloud-Based Knowledge Graphs to Recognize Suicidal Mental Tendencies

Area of Science:

• Cloud Manufacturing
• Knowledge Services
• Psychological Behavior Analysis
• Artificial Intelligence

Background:

• Selecting appropriate knowledge services in cloud manufacturing is challenging.
• User psychological behavior significantly influences service selection and effectiveness.
• Existing methods lack the integration of psychological insights for optimization.

Purpose of the Study:

• To propose a cloud manufacturing knowledge service optimization decision method based on users' psychological behavior.
• To develop a system for identifying psychological issues and suicidal tendencies.
• To enhance the accuracy and efficiency of psychological counseling knowledge acquisition and danger prediction.

Main Methods:

• Established an optimal evaluation index system for cloud manufacturing knowledge services.
• Utilized rough set theory for initial weight assignment and adjusted weights based on user multiattribute preference.
• Developed a knowledge graph from psychological data, employing Han language processing for word segmentation and Chi-square (CHI) feature selection for classification.
• Implemented a bidirectional long short-term memory (BiLSTM) model for detecting suicidal tendencies and a text classifier for analyzing user utterances.

Main Results:

• The proposed system demonstrated superior performance in accuracy, recall rate, and F1 value compared to other standard models in detecting psychological issues.
• The system effectively answers questions related to psychological counseling and identifies users with suicidal tendencies.
• Knowledge graph construction and CHI feature selection proved effective for problem classification and keyword generation.

Conclusions:

• The user psychological behavior-based optimization method significantly improves cloud manufacturing knowledge service selection.
• The developed system offers a reliable tool for counselors to acquire psychological knowledge and identify at-risk individuals.
• The integration of advanced NLP and machine learning techniques provides a robust solution for psychological issue detection.