Skripsi

ANALISIS KOMPARATIF MACHINE LEARNING DAN EXPLAINABLE ARTIFICIAL INTELLIGENCE PADA PREDIKSI FATALITAS KECELAKAAN LALU LINTAS

---

Traffic accidents are a road safety issue that can result in fatalities. This study aims to compare the performance of machine learning models namely, Random Forest, XGBoost and LightGBM and to explain the prediction results of the best model using the Explainable Artificial Intelligence (XAI) approach, with SHapley Additive exPlanations (SHAP) employed as the interpretation method. The data used were sourced from the Brazilian Federal Highway Police (Polícia Rodoviária Federal) for the period 2020–2024. The research stages included data preprocessing, the creation of binary target classes (fatal and non-fatal), data splitting using an 80:20 random split, and validation of the training data using Repeated K-Fold Cross-Validation with K=10 across 3 runs. Evaluation was performed using accuracy, precision, recall, F1-score, ROC-AUC, and a confusion matrix. The results showed that Random Forest performed best overall, with an accuracy of 0.93 and an ROC-AUC of 0.95. However, LightGBM was selected as the best model because it was more aligned with the focus on fatalities, specifically achieving a recall of 0.79, the highest True Positive rate, and the lowest False Negative rate. SHAP analysis indicated that the most influential features were vehicle_type, road_type, and accident_type. These findings support the selection of a more targeted and accurate model.

---

Availability

Detail Information

Series Title

-

Call Number

T2005642026

Publisher

Indralaya : Prodi Sistem Komputer, Fakultas Ilmu Komputer Universitas Sriwijaya., 2026

Collation

xiv, 98 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

006.307

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Sistem Komputer
Model Deep Learning --AI

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment