❮ Previous

Deployment Basics

  • This lesson explains how machine learning models are deployed into production systems and used in real applications.

  • Model Saving

    Once training is complete, we save the model.

    Save Model (TensorFlow / Keras)

    Recommended Format (.keras)

    model.save("my_model.keras")


    This saves:

    • Architecture

    • Weights

    • Optimizer state

    Older Format (.h5)

    model.save("my_model.h5")

    Load Model Later

    from tensorflow.keras.models import load_model

    model = load_model("my_model.keras")

    Now model is ready for predictions.


    Basic Deployment Idea

    Deployment Architecture:

    User → Web/App Interface → Backend Server → Model → Prediction → User

    Step-by-Step Deployment Flow

    User enters input (form/image/text)
    Backend receives data
    Preprocess input
    Load trained model
    Generate prediction
    Return result to user


    Simple Local Deployment Idea (Streamlit)

    Example: Deploy ANN Model

    Step 1: Install Streamlit

Build a Customer Churn Prediction Web App using Streamlit and TensorFlow Keras

This Python example demonstrates how to create a web application for predicting customer churn using Streamlit and a pre-trained TensorFlow Keras model. The code loads the trained model, takes user inputs through the Streamlit interface, makes predictions, and displays whether the customer is likely to churn or stay. The app can be run locally using streamlit run app.py. 

pip install streamlit

Step 2: Create app.py
import streamlit as st
import numpy as np
from tensorflow.keras.models import load_model

# Load model
model = load_model("my_model.keras")

st.title("Churn Prediction App")

# Example inputs
feature1 = st.number_input("Tenure")
feature2 = st.number_input("Monthly Charges")

if st.button("Predict"):
    data = np.array([[feature1, feature2]])
    prediction = model.predict(data)

    if prediction > 0.5:
        st.write("Customer will Churn")
    else:
        st.write("Customer will Stay")


Step 3: Run App
streamlit run app.py

  • Browser opens automatically.


    Types of Deployment

    Type

    Example

    Local App

    Streamlit

    Web App

    Flask / Django

    API

    FastAPI

    Cloud

    AWS, Azure, GCP

    Mobile

    TensorFlow Lite

    Edge Device

    IoT devices


    Basic Deployment Architecture

    Local Deployment

    Model runs on your laptop.

    API Deployment

    Frontend → API (Flask/FastAPI) → Model → JSON Response


    Important Deployment Considerations

    Save preprocessing steps (scaler, encoder)
    Use same feature order
    Handle missing values
    Secure API
    Monitor performance

    Example: Saving Scaler

    import joblib

    joblib.dump(scaler, "scaler.pkl")

    Load later:

    scaler = joblib.load("scaler.pkl")

    Real-World Deployment Flow

    Train Model

         ↓

    Save Model + Scaler

         ↓

    Create API

         ↓

    Test Locally

         ↓

    Deploy to Cloud

         ↓

    Monitor Performance

❮ Previous