Energy Consumption Prediction

A full-stack project for exploring, modeling, and serving energy consumption predictions using Jupyter notebooks, a Python FastAPI backend, and a Next.js frontend.

Quick links

• Backend entry: backend/app.py
• Frontend app: frontend/app/page.tsx
• Dataset: data/raw/Energy_consumption.csv
• Notebooks: notebooks/01_eda.ipynb, notebooks/02_Preprocessing.ipynb, notebooks/03_Modeling.ipynb, notebooks/4_Analysis_Results.ipynb

Repository overview

• backend/ — FastAPI server and model-loading code (backend/app.py).
• frontend/ — Next.js UI and components.
• data/raw/ — Raw CSV dataset (data/raw/Energy_consumption.csv).
• notebooks/ — EDA, preprocessing, modeling, and analysis notebooks.
• models/ — Trained model artifacts (save model.pkl here).
• requirements.txt — Python dependencies.

Prerequisites

• Python 3.9+ and a virtual environment manager (venv/virtualenv).
• Node.js 18+ and npm (for frontend).

Backend — run the API

1. Create and activate a Python environment, then install dependencies:

python3 -m venv myenv
source myenv/bin/activate
pip install -r requirements.txt

2. Start the API (uses uvicorn) — the FastAPI app is in backend/app.py:

uvicorn backend.app:app --reload --host 127.0.0.1 --port 8000

3. Open interactive docs (if running) at http://127.0.0.1:8000/docs.

API endpoints (examples)

The backend exposes the following endpoints (see backend/app.py for implementation details):

• GET / — root welcome message
• GET /health — model health/status
• POST /predict — predict energy consumption; accepts a JSON body with feature values

Health check (cURL)

curl -X GET http://127.0.0.1:8000/health

Predict (cURL)

Example request payload for POST /predict — fields match the InputFeatures Pydantic model in the API:

{
  "Temperature": 24.5,
  "Humidity": 45.0,
  "SquareFootage": 1200.0,
  "Occupancy": 3.0,
  "RenewableEnergy": 0.2,
  "HeatIndex": null,
  "HVACUsage": "On",
  "LightingUsage": "On",
  "DayOfWeek": "Monday",
  "Holiday": "No"
}

Use this cURL command to post a prediction request:

curl -s -X POST http://127.0.0.1:8000/predict \
  -H "Content-Type: application/json" \
  -d '{"Temperature":24.5,"Humidity":45.0,"SquareFootage":1200.0,"Occupancy":3.0,"RenewableEnergy":0.2,"HeatIndex":null,"HVACUsage":"On","LightingUsage":"On","DayOfWeek":"Monday","Holiday":"No"}'

Example successful response:

{ "prediction": [123.45] }

Notes about the model

• The app expects a serialized model at models/model.pkl. The application attempts to load the model on startup and exposes model error details via /health if loading fails (see backend/app.py).
• If HeatIndex is omitted or null, the API computes a simple fallback: HeatIndex = Temperature + 0.5 * Humidity.

Frontend — run the web UI

cd frontend
npm install
npm run dev

Open http://localhost:3000 to view the UI.

Makefile & Docker

This repo includes a Makefile and docker-compose.yml to simplify local development and containerized runs.

Makefile (common targets):

• make up: Build and start services with Docker Compose (uses docker-compose.yml).
• make build: Build backend and frontend containers.
• make backend: Start only the backend locally (virtualenv).
• make frontend: Start only the frontend (npm).
• make down: Stop and remove containers.
• make logs: Show combined logs.
• make clean: Remove built images and temporary artifacts.

Examples:

# start with Docker Compose
make up

# run backend locally in venv
make backend

# run frontend locally
make frontend

# stop containers
make down

Docker usage (docker-compose):

A docker-compose.yml is provided for containerized development. To build and run:

docker-compose build
docker-compose up -d
docker-compose logs -f

The backend exposes port 8000 (FastAPI) and the frontend exposes port 3000 by default — adjust ports in docker-compose.yml as needed.

Data & notebooks

• Reproduce analysis and model training with the notebooks in notebooks/ (recommended order: EDA → Preprocessing → Modeling → Analysis results).
• Store final trained artifacts in models/ and ensure backend/app.py can load models/model.pkl.

Development & improvements

• Add a LICENSE and .gitignore to exclude myenv/ and large artifacts.
• Add unit tests for preprocessing and API endpoints.
• Consider versioning models (MLflow or timestamped filenames) and adding CI pipelines.

---