Skip to content

Fleet Decision Platform

  • Set up in 5 minutes

    Get up and running with the Fleet Decision Platform quickly using our streamlined installation process.

    Getting Started

  • Highly Configurable

    Customize every aspect of the platform through YAML configuration files without changing code.

    Configuration

  • Demand Forecasting

    Multi-location time-series forecasting with XGBoost, Prophet, and hierarchical reconciliation.

    Forecasting Guide

  • Fleet Optimization

    Cascading optimization with min-cost flow and MILP for cost-effective fleet allocation.

    Optimization Guide

Overview

The Fleet Decision Platform is an enterprise-grade decision intelligence system that transforms demand forecasts, contract rules, and risk predictions into explainable, cost-optimized fleet allocation decisions.

graph LR
    A[Demand Data] --> B[Forecasting]
    B --> C[Optimization]
    D[Fleet State] --> C
    E[Constraints] --> C
    C --> F[Allocation Plan]
    F --> G[KPIs & Explanations]

Key Capabilities

Capability Description Phase
Demand Forecasting Multi-location time-series with XGBoost/Prophet MVP
Fleet Optimization Min-cost flow + MILP cascading optimization MVP
Risk Prediction Asset failure and violation prediction Phase 2
Contract Intelligence NLP extraction from contract PDFs Phase 3
Explainability SHAP analysis and cost drivers Phase 2

Quick Start

# Clone the repository
git clone https://github.com/yourusername/fleet-cascade.git
cd fleet-cascade

# Install dependencies
uv sync

# Run the API
make run

import httpx

response = httpx.post(
    "http://localhost:8000/api/v1/optimize",
    json={
        "demand_forecast": {...},
        "fleet_state": {...},
        "constraints": {...}
    }
)
result = response.json()

# config/config.yaml
forecasting:
  model: "xgboost"
  horizon_days: 7

optimization:
  solver: "ortools"
  constraints:
    max_distance: 100

Who Is This For?

Business Stakeholders
Understand fleet optimization outcomes, KPIs, and business impact through intuitive dashboards and reports.
Developers
Build and extend the platform with our modular, well-documented codebase and API.
Operations Teams
Deploy, monitor, and maintain the platform in production environments.
Data Scientists
Experiment with forecasting models, optimization algorithms, and feature engineering.

Architecture at a Glance

graph TB
    subgraph Data["Data Layer"]
        NYC[NYC Taxi Data]
        NASA[NASA Turbofan]
        SIM[Simulated Fleet]
    end

    subgraph ML["ML Layer"]
        FORE[Demand Forecasting]
        RISK[Risk Scoring]
    end

    subgraph OPT["Optimization"]
        CASCADE[Cascading Optimizer]
    end

    subgraph API["API Layer"]
        FAST[FastAPI]
    end

    Data --> ML
    ML --> OPT
    OPT --> API

Project Status

Current Phase: MVP (Phase 1)

The platform is currently in MVP development, focusing on:

  • Project structure and configuration
  • Data ingestion pipeline
  • XGBoost demand forecasting
  • Min-cost flow optimization
  • Basic API endpoints
  • Integration testing

Getting Help

  • Open Source

    Licensed under MIT. Free to use, modify, and distribute.

  • Actively Maintained

    Regular updates and improvements based on community feedback.