PERISAI (Non-Communicable Disease Risk Prediction)

PERISAI is a Fullstack Web Application system based on Artificial Intelligence (AI) designed to analyze lifestyle and patient medical records to predict the risk of 3 major Non-Communicable Diseases (NCDs): Diabetes, Hypertension (HighBP), and High Cholesterol (HighChol).

This system is driven by a custom Deep Learning model (TensorFlow/Keras) trained using a Functional API architecture, Custom Layers, and a Custom Loss Function (BCE + 10x MAE) to generate highly precise probability levels based on real data (BRFSS Dataset). Equipped with a Generative AI assistant for interactive medical consultation.

---

🏗️ Project Architecture & Deployment

This project has a Production-Ready status and is deployed using a modern Microservices architecture distributed decentrally across three different Cloud platforms to ensure scalability and workload isolation.

1. Frontend UI (Vercel)

• Tech Stack: React, Vite, and TailwindCSS.
• Role: Provides the visual interface (Dashboard, AI Prediction Form, Daily Habit Log, Interactive Chatbot).
• Live URL: https://perisai-ptm.vercel.app/

2. Backend API (Railway)

• Tech Stack: Node.js, Express, and PostgreSQL.
• Role: Handles User management (JWT Authentication), Habit Log storage, and acts as an API Gateway bridging the Frontend with the AI Microservice. Autonomously calculates Historical Moving Average (last 7 days average) before sending data to the AI.
• Live URL: https://perisai.up.railway.app

3. AI Microservice (Hugging Face Spaces)

• Tech Stack: Python, FastAPI, TensorFlow, Google Gemini API, Docker.
• Role: High-performance server dedicated exclusively to running Deep Learning model inference (perisai_model_production.keras) and processing Generative AI responses in real-time. Isolated using a Docker Container.
• Live URL: https://huggingface.co/spaces/hilmyinaja/perisai-ai-api

---

📂 Unified Repository Structure (Monorepo)

• /ai_engineer: The heart of Data Science. Contains Jupyter Notebooks for model training, Class Imbalance visualization, TensorBoard logs, and FastAPI API source code.
• /Fullstack/Frontend: Source code for the web user interface.
• /Fullstack/Backend: Source code for Node.js server logic and database management.
• /data: Storage for raw and cleaned datasets (BRFSS CDC).

---

🚀 How to Run the Project (Local Development)

You need to run all three services simultaneously in different terminals if you want to run this application locally.

Prerequisites:

• Python 3.10+
• Node.js 18+
• PostgreSQL Server

Step 0: Clone the Repository

Open a terminal and download the source code from GitHub: bash git clone https://github.com/hilmyinaja/Data-Analysis-Penyakit-Tidak-Menular.git cd Data-Analysis-Penyakit-Tidak-Menular

Step 1: Database & Backend Setup (Node.js)

1. Open a terminal and navigate to the Backend directory:

   cd Fullstack/Backend

2. Install the required libraries:

   npm install

3. Create a .env file inside the Backend folder containing your PostgreSQL URL, for example:

   DATABASE_URL=postgres://postgres:your_password@localhost:5432/perisai_db
   PORT=5000
   JWT_SECRET=super_safe_secret_123
   ML_SERVICE_URL=http://127.0.0.1:8001/api/v1
   AI_SERVICE_URL=http://127.0.0.1:8001/api/v1
   AI_CHAT_URL=http://127.0.0.1:8001

4. Migrate the tables to the database:

   npm run migrate

5. Run the Backend server:

   npm run dev

Step 2: Set Up Python Environment

1. Open a terminal in the root folder of the project.
2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install all dependencies:

   pip install -r requirements.txt

Step 3: Run AI Microservice (FastAPI)

1. Open a new terminal (make sure venv is active) and navigate to the AI API directory:

   cd ai_engineer/perisai_api

2. Create an .env file in that folder and fill it with your Google API Key:

   GEMINI_API_KEY=YOUR_SECRET_KEY

3. Run the FastAPI server:

   uvicorn perisai_api:app --port 8001 --reload

Step 4: Run Frontend (React UI)

1. Open a new terminal and navigate to the Frontend directory:

   cd Fullstack/Frontend

2. Install dependencies:

   npm install

3. Create a .env file containing the Node.js Backend address:

   VITE_API_URL=http://localhost:5000/api

4. Start the web interface:

   npm run dev

5. Open the local link that appears (usually http://localhost:5173) in your browser.

---

📷 Video Documentation

PERISAI.Video.Demo.1.mp4

---

This project is an industry-level demonstration of the intersection of Data Science, Machine Learning Engineering, and Modern Web Software Engineering (Fullstack Development).