Name: R.
Mohan Krishna
Ms. Poorani.S
Register Number: 192211555
Approval
Guided by Dr. Mary No: SSE/26/11/555-02
Valantina. G
Guided by Dr.R Ramya
Enhancing Road Safety a Comparative Analysis of Accident Detection using Random Forest and K Nearest
Neighbour
INTRODUCTION
➢ Develop and compare a Random Forest (RF) with a K Nearest Neighbour (KNN) for an enhanced Road Accident data
classification model, focusing on accuracy, precision, and recall metrics.
➢ This study evaluates the effectiveness of Random Forest (RF) versus K Nearest Neighbour (KNN) for enhancing Road
Accident data classification accuracy, aiming to improve patient care and outcomes in Road Accident management.
➢ This study holds significance in improving Road Accident data classification by comparing the Random Forest (RF) and the K
Nearest Neighbour (KNN). Insights gained can enhance clinical decision-making, inform future research, and drive
advancements in Road Accident management practices.
➢ Our study demonstrates the effectiveness of Random Forest (RF) and K Nearest Neighbour (KNN) in enhancing Road
Accident data classification. RF emphasises sequential ensemble learning for heightened accuracy, while KNN leverages
neural network architecture for intricate pattern recognition, enriching classification outcomes.
MATERIALS AND METHODS
Frame Works High dimensional Low dimensional
Pre-Processing /Libraries: Feature Vectors Feature Vectors
Labeled Data Data 1.Python Language
2.Scikit - learn 3.R
Data Processing and Language Feature Selection
Classification Algorithm
Dataset Feature Extraction Hyper parameters:
1. ROC Curves
2.Confusion Matrices
Un-Labeled
Data High dimensional
Low dimensional
Feature Vectors
Feature vectors
Enhancing of
New Review Data Processing and Training and Prediction Stage Prediction Model Road
Feature Extraction Accident
Enhancing of Road Accident
RESULTS
Groups N Mean Std. Deviation Std.Mean Error
RF 10 0.9250 0.15811 0.05000
Accuracy
KNN 10 0.8289 0.01507 0.00477
Fig:1 Random Forest (RF) is compared with K Nearest
Neighbour (KNN) it depicts that the proposed Table 1: Group Statistics Results –RANDOM FORESTAND DECESION TREE
algorithm(RF) gives more accuracy when compared with
the existing algorithm(KNN).
DISCUSSION AND CONCLUSION
➢ The usefulness of the RF algorithm is demonstrated by its remarkable 92% accuracy rate in detecting the Road Accident . The KNN algorithm, in contrast, does remarkably
well, attaining an amazing 82.89% accuracy. The statistical analysis confirms the significance of both procedures with p = 0.000 (p <0.05).
➢ To enhance Road Accident detection, future endeavours may entail exploring hybrid models amalgamating the strengths of RF with K Nearest Neighbour aiming for heightened
accuracy and robustness. Additionally, advancements could focus on crafting tailored architectures tailored to handle the nuanced characteristics inherent in enhancing road
safety, thereby refining diagnostic precision and efficacy.
➢ One limitation is the potential computational complexity associated with training models like RF, which may demand substantial resources. Moreover, K Nearest Neighbour
Algorithm might encounter challenges accurately capturing intricate patterns in Road Accident data, potentially impacting its comparative performance in accuracy
assessment.
➢ As a result, our research indicates that RF outperforms K Nearest Neighbour Algorithm in detecting intricate Road Accident patterns. While both provide insights, RF shows
promise for more precise classification, enhancing diagnostic accuracy and treatment efficacy.
➢ Comparison of Random Forest and K Nearest Neighbour Algorithm performance. Analysis of algorithm strengths and weaknesses. Evaluation of hyper-parameters and model
tuning effects. Examination of reasons behind observed performance differences. Exploration of unexpected findings and challenges.
➢ Recap of objective: designing improved Road Accident data classification model. Summary of key findings on accuracy, precision, recall. Declaration of superior algorithm.
Implications for healthcare practitioners, researchers, patients. Suggestions for future research Importance of advanced ML for Road Accident data classification.
BIBLIOGRAPHY
1. Kastrinaki, V., & Chronopoulos, G. (2012). A comparative analysis of machine learning techniques for traffic accident prediction. Expert Systems with Applications, 39(5),
6007-6019.
2. Priyadarsini, R., Sankar, A., & Durai, A. (2017). A comparative study of K-nearest neighbor, artificial neural network, decision tree and random forest classifier for accident
prediction. International Journal of Control Theory and Applications, 10(21), 103-110.
3. Goh, C. F., & Loh, S. F. (2017). An investigation of the use of random forests for traffic sign recognition. Procedia Computer Science, 105, 90-95.
4. Hastie, T., Tibshirani, R., & Friedman, J. (2009). The Elements of Statistical Learning: Data Mining, Inference, and Prediction. Springer Science & Business Media.
5. Bishop, C. M. (2006). Pattern Recognition and Machine Learning. Springer. Zheng, J., Yuan, H., & Xu, L. (2015).
6. Comparative Study of Accident Detection Models Based on Decision Trees, Neural Networks, and Hybrid Models. Mathematical Problems in Engineering, 2015, 1-12.
7. Lv, J., Liu, S., & Su, Y. (2019). Comparative study of traffic accident detection based on different algorithms. IOP Conference Series: Earth and Environmental Science, 357(4),
042031.
8. Vaitla, A. D., & Karthikeyan, K. (2017). A Comparative Analysis of Classification Techniques for Road Traffic Accident Data. Procedia Computer Science, 115, 431-438.
9. Marsland, S. (2009). Machine Learning: An Algorithmic Perspective. CRC Press. Murphy, K. P. (2012). Machine Learning: A Probabilistic Perspective. MIT Press.
