A meta fusion model combining geographic data and twitter sentiment analysis for predicting accident severity

Area of Science:

• Artificial Intelligence
• Transportation Engineering
• Data Science

Background:

• Deep learning and real-time data processing have advanced traffic management.
• Integrating diverse data sources remains a challenge for accurate accident prediction.

Purpose of the Study:

• To develop an innovative approach for traffic accident prediction using integrated data.
• To enhance prediction accuracy by combining traditional traffic data with social media insights.

Main Methods:

• Developed ConvoseqNet, a hybrid model combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks.
• Integrated geographic data and Twitter sentiment analysis with traffic data.
• Proposed MetaFusionNetwork, a meta-model combining ConvoseqNet and Random Forest Classifier predictions.

Main Results:

• ConvoseqNet achieved the highest predictive accuracy, effectively capturing accident-related patterns.
• MetaFusionNetwork demonstrated the benefits of ensemble methods for improved prediction.
• The study highlights advancements in real-time data-driven traffic management.

Conclusions:

• The proposed approach significantly advances traffic accident prediction accuracy.
• Leveraging heterogeneous data sources and fusion techniques enhances road safety.
• The research offers insights into model interpretability and computational efficiency.