Abstract :

Traffic Accidents Are A Leading Cause Of Mortality And Property Damage Globally. The Ability To Predict The Severity Of An Accident Based On Environmental And Road Conditions Can Significantly Improve Emergency Response And Traffic Management. Traditional Traffic Safety Models Are Largely Reactive, Relying On Post-accident Analysis, Or They Utilize Standalone Machine Learning Algorithms That Fail To Handle The Severe Class Imbalance Present In Real-world Crash Data (where Minor Accidents Vastly Outnumber Fatal Ones). To Address These Challenges, This Study Proposes A Novel Hybrid Artificial Intelligence System Known As RFCNN, Which Integrates Machine Learning And Deep Learning Techniques Through Decision-level Fusion. The Proposed Approach Utilizes A Comprehensive Road Safety Dataset Containing Heterogeneous Features Such As Weather Conditions, Light Conditions, Road Surface Type, Speed Limits, And Time Of Day. Data Preprocessing Steps—including Handling Missing Values, Label Encoding, And Z-score Scaling—are Applied To Standardize The Inputs. Crucially, The Synthetic Minority Over-sampling Technique (SMOTE) Is Employed To Resolve Class Imbalance, Geometrically Synthesizing Fatal Accident Records So The Models Can Prioritize Severe Crashes Effectively. Central To This Architecture Is A Parallel Processing Framework. The Data Is Fed Simultaneously Into A Random Forest (RF) Model To Extract Structured, Rule-based Decisions, And A 1-Dimensional Convolutional Neural Network (1D-CNN) To Capture Complex, Non-linear Relationships And Latent Spatial Correlations In The Feature Vector. The Predictions From Both Models Are Fused Using A Soft-voting, Weighted Averaging Strategy. Extensive Experiments Demonstrate That The Hybrid RFCNN Model Significantly Outperforms Baseline Approaches, Achieving An Overall Accuracy Of 86.42% And Raising Fatal Accident Recall From 15% To 86%. This Work Contributes To The Field Of Intelligent Transportation Systems (ITS) By Offering A Robust, Deployable Real-time Risk Assessment Tool Via A Stream Lit Web Interface.

Published:

21-7-2026

Issue:

Vol. 26 No. 7 (2026)

Page Nos:

1048-1057

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite