# Data Modeling with Apache Cassandra

**Udacity Data Engineering Nanodegree — Project 2**

## Overview

Model and query a NoSQL database using **Apache Cassandra** for the fictional Sparkify music streaming service. Unlike relational databases, Cassandra requires a _query-first_ design approach: each table is purpose-built to answer one specific analytical question.

## Dataset

Raw event data is sourced from CSV log files (`event_data/2018-11-*.csv`) representing daily user activity. These files are pre-processed and merged into a single consolidated dataset (`event_datafile_new.csv`) before loading into Cassandra.

**Sample fields:** `artist`, `firstName`, `gender`, `itemInSession`, `lastName`, `length`, `level`, `location`, `sessionId`, `song`, `userId`

## Schema Design

Three tables are modeled to answer three specific queries:

| Table | Partition Key | Clustering Columns | Answers |
|---|---|---|---|
| `session_songs` | `sessionId` | `itemInSession` | What song was played in a given session and item? |
| `user_session_songs` | `userId`, `sessionId` | `itemInSession` | What songs did a user listen to in a session? |
| `song_listeners` | `song` | `userId` | Who listened to a specific song? |

## Key Concepts

- **Query-first modeling** — schema designed around queries, not entities
- **Denormalization** — data is duplicated across tables to enable fast reads
- **Partition keys** — determine data distribution across nodes
- **Clustering columns** — control sort order within a partition
- **`cassandra-driver`** — Python client for Cassandra

## How to Run

1. Install dependencies:
   ```bash
   pip install cassandra-driver pandas
   ```

2. Open and run the notebook:
   ```bash
   jupyter notebook "Project_1B_ Project_Template.ipynb"
   ```

The notebook walks through preprocessing the CSV data, creating Cassandra tables, inserting records, and running validation queries.