https://github.com/xueeinstein/social-engagement-predictor

https://github.com/xueeinstein/social-engagement-predictor

Last synced: 1 day ago
JSON representation

• Host: GitHub
• URL: https://github.com/xueeinstein/social-engagement-predictor
• Owner: xueeinstein
• Created: 2014-07-18T08:11:10.000Z (over 10 years ago)
• Default Branch: master
• Last Pushed: 2014-09-12T12:26:53.000Z (about 10 years ago)
• Last Synced: 2024-04-16T03:31:15.776Z (7 months ago)
• Language: Python
• Size: 6.89 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Social Engagement Predictor

## About

This is a project for [RecSys Challenge 2014](http://2014.recsyschallenge.com). However, it has already out of date before I finish this project.

## Main Idea

Not like other traditional collaborative filtering recommendation system, I try to find the real connection relationship by mining the [MovieTweeting](https://github.com/xueeinstein/xueeinstein.github.com/blob/master/assets/doc/crowdrec2013_Dooms.pdf) dataset. And then tracks the Twitter user community hot engagement. According to this result, model and predict engagement ranking list.

## Review

Because of the large dataset, I couldn't handle it directly on my 4G memory laptop :( . So I built a tedious system. Firstly, import the dataset into MongoDB, and the using MapReduce which integrates with MongoDB grouped and filtered the useful data. To handle a two hundred thousand nodes and millions of relationship tweets graph, I chose [Neo4j](http://www.neo4j.org). It is awesome, especially its cypher language. 

Now, I'm trying [gephi](https://gephi.org) to visualize it.