• Krishnakanth Gunta
• Satya Sai Baba Mudigonda
• Phani Krishna Kandala
• Rohan Yashraj Gupta
• Pallav Kumar Baruah

Automobile crashes are a worldwide problem that result in deaths and injuries, in addition to some direct and indirect expenses. Predicting the risk of accidents can be useful for an insurance company to better estimate claims costs from such accidents. The majority of this research looks at the effects of other vehicles, the environment, and weather on crashes. The data for this study is taken from the Waka Kotahi Crash Analysis System (CAS), which tracks all road accidents recorded by the New Zealand Police. The data is preprocessed by handling missing values and tuning the features post which descriptive analysis is performed. Additionally, to handle data imbalance of fatal injuries, the synthetic minority oversampling technique (SMOTE) has been performed. An artificial neural network (ANN) model is then used to classify a crash as a fatal injury or not. The Artificial neural network model is then applied to both the unbalanced & balanced datasets. Their predictive accuracy, precision, recall, F1 score, and area under the receiver operator characteristic (AUC-ROC) have been compared. We thereby demonstrate that the artificial neural network model with balanced data gives the best accuracy (80%), precision (81%), recall (81%), f1-score (81%), and AUC-ROC (.80). Since artificial neural networks are black-box models, local interpretable model agnostic explanations (LIME) is used to check the model interpretability.

Automobile crashes, Fatal injury, Insurance, Data imbalance, Crash Analysis System, Artificial neural network, synthetic minority oversampling technique, local interpretable model agnostic explanations.

"BENEFITS OF USING SMOTE ALONG WITH NEURAL NETWORKS FOR AUTOMOBILE FATAL INJURY CLASSIFICATION MODELS", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.9, Issue 6, page no.j793-j803, June-2022, Available :http://www.jetir.org/papers/JETIR2206999.pdf

"BENEFITS OF USING SMOTE ALONG WITH NEURAL NETWORKS FOR AUTOMOBILE FATAL INJURY CLASSIFICATION MODELS", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.9, Issue 6, page no. ppj793-j803, June-2022, Available at : http://www.jetir.org/papers/JETIR2206999.pdf