---
title: CineMind AI
emoji: 🎬
colorFrom: blue
colorTo: indigo
sdk: gradio
app_file: app.py
pinned: false
sdk_version: 6.18.0
---

# CineMind AI Dataset Information

## Overview

This project is built using the Netflix Prize dataset, a large-scale benchmark dataset widely used for recommendation system research.

The original dataset contains:

* 100,480,507 ratings
* 480,189 anonymous users
* 17,770 movie titles
* Ratings collected between October 1998 and December 2005
* Rating scale: 1–5 stars

To comply with licensing requirements and reduce deployment size, the full Netflix dataset is **not included** in this repository.

Instead, this project uses processed and derived artifacts generated from the original dataset for recommendation inference and demonstration purposes.

---

## Project Purpose

CineMind AI demonstrates the application of modern recommendation system techniques using:

* DeepFM (Deep Factorization Machines)
* NeuMF (Neural Matrix Factorization)
* SVD++

The platform provides:

* Personalized Movie Recommendations
* Similar Movie Discovery
* Intelligent Movie Search
* Recommendation Analytics
* Model Evaluation Dashboard

---

## Data Processing Pipeline

Original Netflix Dataset

↓

Data Cleaning

↓

Feature Engineering

↓

Model Training

↓

Recommendation Generation

↓

Deployment Artifacts

---

## Models Evaluated

### DeepFM

Combines:

* Factorization Machines
* Deep Neural Networks

Captures both low-order and high-order feature interactions.

### NeuMF

Neural collaborative filtering architecture for user-item interaction learning.

### SVD++

Matrix factorization approach incorporating implicit feedback.

---

## Evaluation Results

| Metric     | DeepFM | SVD++ | NeuMF |
| ---------- | ------ | ----- | ----- |
| RMSE       | 0.969  | 0.981 | 0.986 |
| MAE        | 0.770  | 0.771 | 0.780 |
| HitRate@10 | 0.457  | 0.318 | 0.312 |
| NDCG@10    | 0.268  | 0.166 | 0.162 |
| MAP@10     | 0.210  | 0.120 | 0.116 |

DeepFM was selected as the final deployment model due to superior recommendation performance.

---

## Repository Contents

This repository contains:

* Trained DeepFM model
* Encoders and preprocessing artifacts
* Movie metadata
* Application source code
* Deployment configuration

The original Netflix Prize dataset files are not redistributed.

---

## Dataset License Notice

This project uses data derived from the Netflix Prize dataset.

The original Netflix dataset remains subject to its respective usage restrictions and licensing terms.

Users seeking access to the original dataset should refer to the official Netflix Prize documentation and licensing information.

---

## Author

Hardik Gautam

CineMind AI — DeepFM-Based Movie Recommendation System