Skip to content

Latest commit

 

History

History
297 lines (212 loc) · 8.74 KB

README.md

File metadata and controls

297 lines (212 loc) · 8.74 KB

Big Data Recommendation Project - IMDb & Netflix (ENGLISH VERSION)

Project Overview

This project implements a complete end-to-end data pipeline using Big Data technologies. It covers everything from data ingestion, cleaning, formatting, transformation (via Spark), machine learning recommendation, and indexing results into Elasticsearch for dashboard visualization using Kibana.

The domain is focused on movies, with cross-source analysis of IMDb and Netflix data to generate valuable movie recommendations.

Data Pipeline Structure

1. Data Ingestion

  • IMDb: Download title.basics.tsv.gz and title.ratings.tsv.gz from IMDb official dataset site.
  • Netflix: Download netflix_titles.csv from Kaggle using API.
  • Scripts: ingest_imdb_api.py, ingest_netflix_api.py

2. Data Formatting

3. Data Combination

  • Join IMDb (titles + ratings) with Netflix titles via lowercase title match
  • Compute recommendation score: recommend_score = averageRating + log(numVotes + 1)
  • Script: combine_recommend.py

4. ALS-Based Recommendation

  • Use pyspark.ml.recommendation.ALS for collaborative filtering
  • Simulate user-item interactions with generated IDs and constant rating
  • Script: als_recommend.py

5. Elasticsearch Indexing + Kibana Dashboard

  • Index both heuristic and ALS-based results to Elasticsearch
  • Visualize key metrics on Kibana dashboard
  • Scripts: index_elastic.py, index_als.py

6. Full Workflow with Airflow DAG

Project Structure

bigdata/
├── data/
│   ├── raw/                # Original source files
│   ├── formatted/          # Cleaned parquet files
│   ├── combined/           # Final recommendation output
│   └── als/                # ALS model results
├── scripts/                # Python scripts
├── dags/                   # Airflow DAG files
├── dashboards/             # Optional Kibana visual configs
├── requirements.txt
└── README.md

Environment Setup

Python Dependencies (requirements.txt)

python-dotenv>=0.21.0
requests>=2.28.1
pandas>=1.5.3
numpy>=1.23.5
pyspark>=3.3.2
elasticsearch>=7.17.10
apache-airflow==2.7.3
apache-airflow-providers-apache-spark==4.0.1
apache-airflow-providers-elasticsearch==4.2.1
apache-airflow-providers-cncf-kubernetes>=6.0.0
kaggle>=1.5.16
jupyterlab>=3.6.3
ipykernel>=6.19.4

System Dependencies

  • Python >= 3.8
  • Java 11 (required by Spark)
  • Apache Spark 3.3+
  • Elasticsearch 7.x (localhost:9200)
  • Apache Airflow 2.x
  • (Optional) Kibana

Project Highlights

  • Cross-source data integration (IMDb + Netflix)
  • Clean data lake structure (raw / formatted / combined)
  • Machine learning with ALS collaborative filtering
  • Airflow orchestration
  • Elasticsearch + Kibana dashboard integration

How to Run

# Create and activate virtual environment
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Start Elasticsearch
elasticsearch &  # Or via Docker

# Launch Airflow
export AIRFLOW_HOME=~/airflow
airflow db init
airflow standalone

# Manually trigger DAG
airflow dags trigger dual_recommend_pipeline

#Use airflow_start.sh to Start Airflow
bash airflow_start.sh

Useful Links

Submission Checklist

  • README.md: Full documentation in Chinese and English
  • requirements.txt: Full dependency list
  • video.mp4: Demonstration (<10 min)
  • project.zip: Source code archive

Big Data Recommendation Project - IMDb & Netflix (CHINESE VERSION)

项目概要

该项目实现了一个基于大数据技术的完整线性数据流程，从数据采集，数据清洗与格式化，到 Spark 统计/推荐算法处理，最终通过 Elasticsearch 实现可视化指标分析。

该项目以影视为主题，选取 IMDB 和 Netflix 两类影视数据源，展示如何通过跨源分析和机器学习算法实现实际价值的推荐系统。

数据流程模型

1. 数据采集 (Ingestion)

2. 数据格式化 (Formatting)

3. 数据组合 (Combination)

  • 利用 Spark 将 IMDb 标题 + 评分与 Netflix 的视频标题 join
  • 计算素数：recommend_score = averageRating + log(numVotes + 1)
  • 脚本：combine_recommend.py

4. ALS 算法推荐

  • 使用 pyspark.ml.recommendation.ALS 基于 Netflix title 模拟用户行为
  • 随机生成 user_id 和评分，进行返回 top N 推荐
  • 脚本：als_recommend.py

5. Elasticsearch 索引 + Kibana 可视化

  • 将两种推荐结果分别索引到 Elasticsearch
  • Kibana 中构建推荐统计值相关的 Dashboard
  • 脚本：index_elastic.py, index_als.py

6. 全流程引擎: Airflow DAG

文件结构

bigdata/
├── data/
│   ├── raw/                # 原始数据源 (IMDb / Netflix)
│   ├── formatted/          # 清洗后格式化 Parquet 文件
│   ├── combined/           # 组合推荐结果
│   └── als/                # ALS 推荐结果
├── scripts/                # Python 脚本
├── dags/                   # Airflow DAG
├── dashboards/             # Kibana 配置 (JSON 或截图)
├── requirements.txt
└── README.md

环境配置

Python 依赖 (requirements.txt):

python-dotenv>=0.21.0
requests>=2.28.1
pandas>=1.5.3
numpy>=1.23.5
pyspark>=3.3.2
elasticsearch>=7.17.10
apache-airflow==2.7.3
apache-airflow-providers-apache-spark==4.0.1
apache-airflow-providers-elasticsearch==4.2.1
apache-airflow-providers-cncf-kubernetes>=6.0.0
kaggle>=1.5.16
jupyterlab>=3.6.3
ipykernel>=6.19.4

系统依赖:

  • Python >= 3.8
  • Java 11 (用于 Spark 运行)
  • Apache Spark 3.3+
  • Elasticsearch 7.x (default: localhost:9200)
  • Airflow 2.x (LocalExecutor)
  • Kibana (optional)

项目特色点

  • 跨源数据组合: IMDb 分数 + Netflix 内容 join 生成实际价值
  • 环境分层明确: raw/formatted/combined 等数据清清楚楚
  • ALS 算法推荐模型: Spark MLlib 实现用户行为应用
  • Airflow DAG 实现全链连数据运行
  • Elasticsearch + Kibana 结果可视化

如何运行

# 配置虚拟环境 (Python >= 3.8)
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# 启动 Elasticsearch
elasticsearch &  # 或 Docker 启动

# 运行 Airflow DAG
export AIRFLOW_HOME=~/airflow
airflow db init
airflow standalone

# 手动执行 DAG
airflow dags trigger dual_recommend_pipeline

#使用 airflow_start.sh 启动 Start Airflow
bash airflow_start.sh

链接参考

项目提交内容

  • README.md: 详细项目说明
  • requirements.txt: 完整环境依赖
  • video.mp4: 运行视频解说 (<10min)
  • project.zip: 代码打包文件