Fleet-ML-Components

This repository contains the Machine Learning (ML) and AI-driven intelligence modules for the Fleet Time Machine project—a comprehensive fleet monitoring and optimization platform designed to streamline vehicle maintenance, optimize routing, enhance safety, and support intelligent ride assignment. GitHub

Key Challenge: Synthetic Dataset Creation

The most difficult and time-intensive part of this project was creating a synthetic dataset that accurately simulates real-world fleet behavior, including maintenance cycles, risk levels, and fuel efficiency. Achieving realism and balance in the dataset required extensive feature engineering and multiple refinement steps.

Contents Overview

Data Preparation & Datasets

cleaned_dataset.csv

fleet_maintenance_dataset_final.csv and its refined/boosted variations: e.g. _refined.csv, _balanced.csv, _boosted_v2.csv

Various risk, regression, and strategy-labeled datasets (e.g., fleet_risk_classification_dataset.csv, fleet_risk_regression_dataset_balanced.csv)

Data Processing Scripts

feature_selection.py

refine_dataset.py

refine_maintenance_dataset.py

Model Training & Evaluation

initial_model.py

test_model.py

testing_models.py

get_metrics.py

Domain-Specific ML Modules

fuel_efficiency_model.py

predicted_speed.py

risk_level.py

Model Artifacts

fleet_nn_model.keras (trained neural network model)

real_time_predictions.csv (sample output)

Models Evaluation Screenshots: