IDEACT — Intelligent SDG-Oriented Hackathon Advisor

Minor Project | B.Tech CSE (AI & ML) | UPES Dehradun | 2025–26
Mentor: Prof. Aryan Gupta, Department of CSE

IDEACT is a web-based AI platform that takes a hackathon problem statement, maps it to relevant UN Sustainable Development Goals using NLP, and generates structured project ideas with a full execution roadmap.

Most hackathon platforms just list problems. SDG portals give theory. Generic chatbots give filler. IDEACT connects all three into a working decision-support system.

---

What It Does

You paste a problem statement. IDEACT:

1. Preprocesses it — tokenization, stopword removal, lemmatization
2. Extracts top keywords using term frequency
3. Classifies it against all 17 UN SDGs using a trained ML model
4. Finds similar problems from a curated hackathon dataset via cosine similarity
5. Generates 1–5 structured project ideas with features, tech stack, SDG alignment, and India scheme fit
6. Builds a week-by-week execution roadmap (1–12 weeks, configurable)

---

Architecture

frontend/
  index.html          ← Single-file UI (no build step needed)

backend/
  main.py             ← FastAPI app entry point
  routes/
    analyze.py        ← POST /api/v1/analyze
    ideas.py          ← POST /api/v1/ideas
    roadmap.py        ← POST /api/v1/roadmap
  services/
    nlp.py            ← Preprocessing pipeline
    similarity.py     ← TF-IDF cosine similarity index
    recommender.py    ← Idea + roadmap generation engine
  model/
    train_model.py    ← Train the SDG classifier
    predict.py        ← Load model + run predictions
    model.pkl         ← Pre-trained Logistic Regression model
    vectorizer.pkl    ← TF-IDF vectorizer
    label_binarizer.pkl
  data/
    hackathon_dataset.json    ← Curated problem statements
    sdg_knowledge_base.json   ← SDG definitions and keywords

Tech Stack: Python · FastAPI · Scikit-learn · NLTK · React (single-file) · HTML/CSS/JS

---

API Endpoints

Method	Endpoint	What it does
GET	/	API info and available endpoints
GET	/health	Model and dataset load status
GET	/api/v1/sdgs	All 17 SDGs with names
POST	/api/v1/analyze	Preprocess + predict SDGs
POST	/api/v1/ideas	Generate project ideas
POST	/api/v1/roadmap	Generate execution roadmap

---

Setup & Running

1. Clone the repo

git clone https://github.com/Ankitnotnani/IDEACT.git
cd IDEACT

2. Create a virtual environment

python -m venv venv
source venv/bin/activate        # Mac/Linux
venv\Scripts\activate           # Windows

3. Install dependencies

pip install -r requirements.txt

4. Train the model

The pre-trained .pkl files are included. If you want to retrain:

cd backend
python model/train_model.py

5. Start the backend

cd backend
uvicorn main:app --reload --host 0.0.0.0 --port 8000

Backend runs at: http://localhost:8000
Swagger docs at: http://localhost:8000/docs

6. Open the frontend

Just open frontend/index.html in your browser. No build step, no npm, no setup.

Make sure the backend is running first.

---

Sample API Usage

Analyze a problem statement:

curl -X POST http://localhost:8000/api/v1/analyze \
  -H "Content-Type: application/json" \
  -d '{"problem_statement": "Farmers in UP cannot access real-time mandi prices, causing 30% crop loss."}'

Generate project ideas:

curl -X POST http://localhost:8000/api/v1/ideas \
  -H "Content-Type: application/json" \
  -d '{"problem_statement": "Rural students lack access to quality digital education", "top_n": 3}'

Generate a roadmap:

curl -X POST http://localhost:8000/api/v1/roadmap \
  -H "Content-Type: application/json" \
  -d '{"problem_statement": "Air quality monitoring in Tier-2 cities", "duration_weeks": 4}'

---

Team

Name	SAP ID	Role
Ankit Notnani	500124273	AI/ML & Backend
Abhishree Panwar	500119231	Data & Research
Chaitanya Gaur	500122644	Frontend
Kshaunish Singh	500120272	Backend & APIs

Mentor: Prof. Aryan Gupta, Department of CSE, UPES Dehradun

---

SDGs Covered

All 17 UN Sustainable Development Goals — from No Poverty (SDG 1) to Partnerships for the Goals (SDG 17). The classifier supports multi-SDG predictions with confidence scores.

---

License

MIT License — free to use, modify, and distribute with attribution.