Full-Stack Smart Recommendations with ML Backend

This project is a full-stack web application that provides smart recommendations using a machine learning model. The entire pipeline, from data handling to model deployment, is included.

Overview

The core of this project is a machine learning model trained to provide recommendations based on user inputs. This model is served via a REST API built with a Python backend. A user-friendly web interface, built with HTML, CSS, and JavaScript, consumes this API to display the smart recommendations to the end-user.

The project demonstrates a complete end-to-end machine learning workflow:

1. Data Acquisition: Gathering the initial data (e.g., zomatofinal.csv).
2. Data Cleaning & EDA: Processing the data and performing Exploratory Data Analysis (as seen in src.ipynb).
3. Model Training: Experimenting with and training a suitable ML model (e.g., Decision Tree, as saved in modeldt.pkl).
4. Deployment: Deploying the model as part of a full-stack web application using Flask.

---

Tech Stack

• Backend: Python, Flask
• Machine Learning: Scikit-learn, Pandas, NumPy
• Frontend: HTML5, CSS, JavaScript
• Data/Model: Jupyter Notebook (src.ipynb), CSV, Pickle (modeldt.pkl)

---

Project Structure

Here is a brief overview of the key files and directories in this project:

├── static/             # Contains static assets (CSS, JS, images)
├── templates/          # Contains HTML templates for the web app
├── app.py              # Main Flask application file (runs the web server and API)
├── modeldt.pkl         # Pre-trained, serialized machine learning model
├── modeldt.py          # Python script used to train and save the ML model
├── src.ipynb           # Jupyter Notebook for data analysis, cleaning, and model experimentation
├── zomatofinal.csv     # Dataset used for training the model
└── ...

---

How It Works

1. ML Model: The src.ipynb notebook is used to explore the zomatofinal.csv dataset. The data is cleaned, features are engineered, and a machine learning model is trained. The best-performing model is saved as modeldt.pkl.
2. Backend (API): The Flask server (app.py) loads the pre-trained modeldt.pkl model. It exposes one or more API endpoints that receive data from the user (via the frontend). This data is fed into the model to get a prediction/recommendation.
3. Frontend (UI): The user interacts with the HTML pages in the templates directory. JavaScript in the static folder captures user input and sends an API request to the backend. The frontend then receives the recommendation from the API and displays it to the user.

---

How to Run

1. Clone the repository:

   git clone [https://github.com/S4srihari/Smart-Recommendations.git](https://github.com/S4srihari/Smart-Recommendations.git)
   cd Smart-Recommendations

2. Set up a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install dependencies: (You may need to create a requirements.txt file first. Based on the files, you'll at least need Flask and scikit-learn).

   pip install Flask scikit-learn pandas

4. Run the application:

   python app.py