🧠 MLServe - FastAPI ML Model Deployment

🚧 UNDER ACTIVE DEVELOPMENT 🚧

📑 Table of Contents

• 💡 Overview
• ✨ Features
• 🏗️ Architecture
• 🛠️ Installation
  • 1. Clone the Repository
  • 2. Install Dependencies
  • 3. Run the api
  • 4. Test API
• 📡 API Endpoints
  • 🔹 Root Endpoint
  • 🔹 Predict
• 📊 Prediction Charts
  • 📊 AAPL
  • 📊 MSFT
  • 📊 GOOG
  • 📌 Interpreting the Prediction Charts
• 🧠 Machine Learning Model Details
  • 📊 Sample Forecast Visualization
  • 📌 Trend and Seasonality Components
  • 🔍 Understanding the Forecast
• 🚀 Future Enhancements

---

💡 Overview

This project demonstrates how to serve a machine learning model using FastAPI and Docker into production as a RESTful API using FastAPI. The API allows users to send data and receive stock price predictions using Prophet ML model.

✨ Features

• FastAPI-based REST API for serving ML predictions.
• Dockerized deployment for easy scalability.
• Asynchronous request handling for efficient inference.
• Model versioning and logging to track performance.
• CI/CD integration (future enhancement) for automated deployments.

🏗️ Architecture

1. Pretrained ML Model – A trained model is saved and loaded for inference.
2. FastAPI Backend – Exposes RESTful endpoints for making predictions.
3. Docker Containerization – The application runs inside a container for portability.
4. (Optional) Cloud Deployment – Can be deployed on AWS/GCP/Azure using Kubernetes or serverless functions.

🛠️ Installation

1. Clone the repository

git clone https://github.com/yourusername/fastapi-ml-deploy.git
cd fastapi-ml-deploy

2. Install dependencies

pip install -r requirements.txt

3. Run the api

docker-compose up

4. Test API

Open your browser and visit:

http://127.0.0.1:8008/docs

📡 API Endpoints

🔹 Predict

POST /predict/

Request Body:

{
    "ticker": "AAPL"
    
}

Response:

{
    "ticker": "AAPL",
    "forecast": 
    [
          {
              "ds": "2025-02-22T00:00:00",
              "trend": 244.01168239424806,
              "yhat_lower": 232.09980241706867,
              "yhat_upper": 249.23639821076168,
              "trend_lower": 244.01168239424806,
              "trend_upper": 244.01168239424806,
              "additive_terms": -2.6701240826347483, 
              .......
          },
          {
              "ds": "2025-02-23T00:00:00",
              "trend": 244.1718582137145,
              "yhat_lower": 232.24244874958652,
              "yhat_upper": 249.32422233689414,
              "trend_lower": 244.1718582137145,
              "trend_upper": 244.1718582137145,
              "additive_terms": -3.333045887720604,
              ......
          },
          .....
    ]
}

📊 Prediction Charts

Here are sample stock price predictions generated using the Prophet model:

Apple (AAPL)

Microsoft (MSFT)

Google (GOOG)

📊 Interpreting the Prediction Charts

When generating stock price forecasts using Prophet, the visualization includes:

• ⚫ Black Dots (y) → Actual historical stock prices used for training.
• 🔵 Blue Line (ŷ or yhat) → The model’s predicted stock price (yhat).
• 🔹 Shaded Light Blue Area → The confidence interval, representing the expected range of price fluctuations:
  • Upper Bound (yhat_upper) → Maximum expected price.
  • Lower Bound (yhat_lower) → Minimum expected price.

📌 Trend and Seasonality Components: The model also breaks down the stock price into trend and seasonality effects which is covered in the next section.

🧠 Machine Learning Model Details

The model used in this project is Facebook Prophet, which forecasts stock prices based on historical market data. It generates time-series predictions, showing both trends and seasonal effects.

📊 Sample Forecast Visualization

Here’s an example of how the model predicts stock prices over time:

Date	Predicted Price ($)
2025-02-20	190.12
2025-02-21	191.45
2025-02-22	192.78

---

📌 Trend and Seasonality Components

Prophet decomposes the time-series data into three main components:

1. Trend – The overall direction of the stock price over time.
2. Seasonality – Repeating patterns at daily, weekly, or yearly intervals.
3. Holidays/Events – External factors that impact stock prices.

Here’s an example of the trend and seasonal effects visualized for MSFT:

---

🔍 Understanding the Forecast

• The blue line represents the predicted stock price.
• Seasonality plots (daily, weekly, yearly) reveal recurring patterns.

These insights can help traders and analysts make informed investment decisions. 🚀

🚀 Future Enhancements

• ✅ Deploy on Kubernetes using Minikube or cloud providers.
• ✅ Implement batch predictions for processing large datasets.
• ✅ Add authentication & API rate limiting for security.
• ✅ CI/CD pipeline to automate testing & deployment.