Amelia Nguyen

Fuel Consumption Prediction for ICE Powertrain Vehicles

Amelia Nguyen29 May 2026

Benchmarking study of ML-based approach for Real-time Fuel Rate Consumption (L/Hr) Prediction for ICEVs with Random Forest, XGBoost, LightGBM, LSTM, and TCN, training on real ODB-II data collected from 300 vehicles in Ann Arbor within 1 year.

Key achievements

Approach the data leakage issue
Final model reached 13.87% MAPE and 0.3 MAE (L/hr) on real-life dataset (Feature engineering that brings MAPE from ~55% to 13.87%
SHAP for feature importance analysis
Summary Article: https://medium.com/@ameliablog/benchmarking-machine-learning-approach-for-real-world-fuel-consumption-prediction-in-icevs-f657408e888b

Deer-Vehicle Collision Prediction

Amelia Nguyen28 Feb 2026

Built predictive models for real deer collisions data using K-Means clustering, supervised learning (Random Forest, Boosting), and statistical models with feature engineering and resampling techniques (PCA, Bootstrapping, Cross Validation).

Key achievements

Build a clustering map for risk-evaluation.
Use ML tree-based models (random forest, boosting) with feature engineering to tackle overfitting and non-linearity challenge.

Vehicle Crash Frequency and Severity Prediction

Amelia Nguyen09 Dec 2025

Deployed predictive modeling and diagnostics for an integrated vehicle crash dataset using statistical models (linear regression & logistics regression)

Key achievements

Utilized regression diagnostics for features interpretation.
Feature engineering (log-transform and interactive effect) to achieve R-squared = 85.1% for linear regression and AUC = 68% for logistic regression.
A full comparison with current research literature on the topic.

Recommendation Systems on Online Retail

Amelia Nguyen08 Dec 2025

Implemented content-based, user-based and item-based collaborative filtering recommendation systems using PySpark to retrieve product recommendations for customer in an online retail scenario.

Key achievements

Built multiple recommendation approaches in PySpark for customer-product matching.
Achieve RMSE = 0.20, Customer Coverage = 99%, Product Coverage = 50% on average.