Table 1 Literature work on predicting crash severity with the use of ML algorithms.
References | Nature of the Problem | Classification/ Severity Levels | Class Imbalance countermeasures | Approaches Utilized (Optimal model) |
|---|---|---|---|---|
Aboulola O34 | Improving traffic accident severity prediction | Fatal, Serious, Minor, No-Injury | Not | MLP, CNN, LSTM, ResNet, MobileNet, (MobileNet) |
Ren et al.31 | Traffic safety assessment, undivided 2-way highway rail grade crossings | Uninjured, Injured, Killed | Not | XGBoost with SHAP |
Shao et al.35 | Injury severity prediction in automotive crashes using natural language processing | Minor, moderate, Major, Catastrophic, | Not, | XGBoost, LGBM, Cat Boost, RF, SVM, NLP, (XGBoost-NLP) |
Chen et al.20 | Improving crash severity with imbalanced data | Fatal Non-fatal | Yes, SMOTE-NC, RUS, GAN | Binary logit with GAN-RU |
Sattar et al.2 | Road crash injury severity prediction | Non-Sever Severe | Not | GNN, RF, XGBoost, ANN (GNN) |
Li et al.10 | Crash injury severity prediction considering data imbalance | No injury, Possible Inj., Non-incap. Inj., Incap. Inj., Fatal | Yes, RUS, ROS, SMOTE, ADASYN and WGAN-GP | LR, MLP, RF, AdaBoost, XGB, (XGBoost with WGAN-GP) |
Aziz et al.36 | Road traffic Crash severity analysis using DES with IH and RoC | Non-Injury Injury | Cost-sensitive learning with Class weights | DES with static ensemble methods, (KNORAE-CatB) |
Kou et al.11 | Classification of autonomous vehicle crash severity by solving the problems of imbalance | Non injured Injured | Yes, ROSE, SMOTE | RF, MI, XGBoost (RF with SMOTE |
Tahfim et al. 4 | Imbalanced data of crashes Involving Trucks | Minor Injury Major Injury | Yes, (ADASYN, Near Miss, SMOTE-Tomek, Under-sampling) | LR, RF, GBDT, MLP, (GBDT with ADASYN) |
Wang et al.21 | Prediction of MV crashes severity based on random forest optimized by meta-heuristic algorithm | Minor impact, Moderate, impactful, significant Imp. | Yes, SMOTE | RF, BP. SVM, XGBoost, (RF with metaheuristic) |
Yang et al.38 | Crash severity analysis with data enhanced with double layer stacking model | Severe Minor No Injury | Yes, ROS | Stacking LR, RF, KNN, GBDT, Cat Boost, GNB (EnLKtreeGBDT) |
Roudnitski18 | Road crash severity prediction with balanced ensemble models | Non-casualty, Minor, Moderate, Serious, Fatal | Yes, (SMOTE) | RF, XGB, AdaB, LGBM, CatB |
Cheng et al.22 | Crash severity prediction for road determinants based on a hybrid method | Minor, Moderate, Major | Yes, SMOTE | XGBoost, SVM, DT, MLP (XGBoost-SMOTE) |
Ogungbire et al.23 | Effectiveness of Data Imbalance Treatment in Weather-Related Crash Severity Analysis | PDO, Moderate, Severe, | Yes, SMOTE-N, ADASYN-N, | XGBoost, RF |
Aldhari et al.9 | Severity prediction of highway crashes in Saudi Arabia | PDO, Injury Death | Yes, (Ran. over sampling) | RF, XGBoost, LR (XGBoost) |
Obasi et al.27 | Forecasting the severity of road accidents | Serious Non-Serious | Not | NB, RF, LR, ANN, (RF and LR) |
Ahmed et al.19 | Road accident prediction and contributing factors | Fatal, Serious Minor, No-Injury | Yes (SMOTE) | DJ, RF, AdaB, CatB, XGB, LGBM, (RF) |
Megnidio-Tchoukouegou et al.32 | Road traffic accident improvements | Fatal, Serious Slight | Not | DT, XGBoost, Light GBM (DT) |
Perez-Sala et al.15 | Prevention of traffic accidents severity | Slight, Severe, Fatal | Yes, (Borderline SMOTE) | CNN, GNB, SVM, KNN (CNN) |
Niyogisubizo et al.29 | Predicting Crash Injury Severity in Smart Cities | Non-Serious Serious | Not | KNN, XGB, AdaB, GB, (Wide & Deep L) |
Asadi et al.24 | Crash severity in work zone area | PDO Injuries | Yes, (SMOTE-ENN, SPE) | SPE, LGBM, AdaB, Cart, BLR (SPE) |
Mohammadpour et al.16 | Trucks involved in fatal accidents | KA (Fatal + Severe) BC (Less severe) O (PDO) | Yes, (SMOTE) | RF, DT, KNN, MLP, GBDT, SVM, (GBDT, RF) |
Raja et al.8 | Road traffic accident in Oromia special zone | Slight, Serious, Fatal | Yes, SMOTE, Ran. over sampling & Under sampling | BPNN, FFNN, MLPNN, RNN, RBFNN, LSTM, |
Vanishkorn et al.17 | Crash severity classification prediction | No Injury, Minor, Fatal | Yes (SMOTE) | DT, RF, GB, LR, KNN (GB with SHAP) |
Islam et al.28 | Predicting road crash severity and Crash hotspots | Fatal Serious Injuries | Not | RF, XGBoost, LR (RF) |
Kim et al.12 | Developing crash severity models handling class imbalance | PDO, Injury, Fatal | Yes (RUMC) | MNL, OLR, RF, ORF, (Imbalanced ORF) |
Sangare et al.14 | Forecasting road accidents | No Injury, Non-Incap. Injury, Incap. Injury | Yes, (SMOTE) | GMM, SVC |
Chen et al.13 | Mechanism of crashes with Autonomous Vehicles | Un-injured, Injury Crashes | Yes, (SMOTE) | SVM, CART, (XGBoost) |
Fiorentini et al. 3 | Handling imbalanced data in road crash severity prediction | Fatal + Injury PDO | Yes (RUMC) | DT, RF, KNN, LR (RF with RUMC) |
Wahab et al.33 | Severity prediction of motorcycle crashes | Fatal, Hospitalized, Injured, Damage | Not | MLP, PART, CART (CART) |
Chakraborty et al.37 | Traffic crash injury severity | Fatal + Sev (KA), Non-Sev + possible (BC), PDO (O), | Yes, (SMOTE) | DT, RF, XGBoost, DNN |
Liu et al.30 | Intelligent Vehicle Crash Injury Severity | No Inj., Slight Inj., Serious Injury | Not | AdaBoost, GBDT, XGBoost (CSSV-AGX) |
