VIP20
本课程属于：029-AI for CAE Simulation VVUQ & Geometric Deep Learning English Technical Courses · 共 20 门相关课程，组合学习更完整。
查看所属VIP全部课程 →
← 返回所属目录 咨询 / 购买

Crash Injury Prediction Using Random Forest and XGBoost

30 Chapters
View Course Samples Online
01
Crash injury prediction fundamentals
problem definition, injury severity scales, and the role of machine learning in traffic safety.
predictiondefinition
02
Data sources for crash injury prediction
police reports, hospital records, vehicle telemetry, and roadway inventory datasets.
predictionhospital
03
Exploratory data analysis for crash data
distribution analysis, missing value patterns, and class imbalance assessment.
Exploratoryanalysis
04
Feature engineering for crash injury prediction
deriving vehicle kinematics, occupant characteristics, and environmental risk factors.
predictionderiving
05
Handling missing and imbalanced crash data
imputation strategies, resampling techniques, and their impact on model fairness.
Handlingimbalanced
06
Random Forest algorithm for injury severity classification
bagging, feature subsampling, and tree aggregation mechanics.
algorithmseverity
07
XGBoost algorithm for injury severity classification
gradient boosting, regularization, and tree pruning mechanics.
XGBoostalgorithm
08
Baseline modeling with Random Forest
hyperparameter tuning, cross-validation, and performance benchmarking on crash data.
Baselinemodeling
09
Baseline modeling with XGBoost
learning rate, tree depth, and early stopping strategies for injury prediction.
Baselinemodeling
10
Feature importance and interpretation in Random Forest
Gini importance, permutation importance, and partial dependence plots.
importanceinterpretation
11
Feature importance and interpretation in XGBoost
gain, cover, SHAP values, and their use in crash injury analysis.
importanceinterpretation
12
Model evaluation for injury prediction
confusion matrices, precision-recall trade-offs, and cost-sensitive metrics.
evaluationprediction
13
Addressing class imbalance in injury severity prediction
SMOTE, class weighting, and focal loss with tree ensembles.
Addressingimbalance
14
Hyperparameter optimization for Random Forest and XGBoost
grid search, random search, and Bayesian optimization.
Hyperparameteroptimization
15
Cross-validation strategies for crash data
stratified k-fold, group k-fold by crash event, and temporal validation.
Cross-validationstrategies
16
Ensemble comparison and selection
Random Forest vs. XGBoost performance, interpretability, and computational trade-offs.
Ensemblecomparison
17
Case study
predicting injury severity in urban intersection crashes using Random Forest and XGBoost.
predictingseverity
18
Case study
highway crash injury prediction with vehicle telemetry and environmental features.
predictiontelemetry
19
Case study
pedestrian and cyclist injury severity modeling with imbalanced crash datasets.
pedestrianseverity
20
Temporal and spatial validation of crash injury models
concept drift, seasonal effects, and regional generalization.
Temporalvalidation
21
Model calibration and risk scoring
converting tree ensemble outputs into actionable injury risk probabilities.
calibrationconverting
22
Fairness and bias in crash injury prediction
demographic disparities, proxy variables, and mitigation strategies.
Fairnessprediction
23
Deployment considerations for crash injury models
real-time inference, model serialization, and API integration.
Deploymentconsiderations
24
Monitoring and maintaining deployed models
performance drift detection, retraining triggers, and data pipeline updates.
Monitoringmaintaining
25
Advanced feature engineering with Random Forest and XGBoost
interaction features, target encoding, and feature selection.
AdvancedXGBoost
26
Handling high-cardinality categorical features in crash data
target encoding, frequency encoding, and native categorical support in XGBoost.
Handlinghigh-cardinality
27
Uncertainty quantification in injury prediction
prediction intervals, conformal prediction, and ensemble variance.
Uncertaintyquantification
28
Integrating domain knowledge into tree ensembles
monotonic constraints, custom objectives, and physics-informed features.
Integratingensembles
29
Comparative analysis with alternative models
logistic regression, support vector machines, and neural networks as benchmarks.
Comparativeanalysis
30
Capstone project
end-to-end crash injury prediction pipeline with Random Forest and XGBoost, from data ingestion to model interpretation and reporting.
Capstoneend-to-end