Data Analysis and R-Node

Jamie Love

N-Squared Software

Introduction

• Who am I
  • Jamie Love (Stack Overflow, LinkedIn)
  • Technical Director of N-Squared Software
• Data and Analysis. Why this talk?
  • Data is gathered, and provided, in ever-increasing amounts.
  • The Internet makes accessing data easier.
  • Analysis is important to understand and reason about data.
  • Open Source software can help with this.
  • R-Node is (will be) about making one Open Source tool more accessible.

Data Analysis

• Massive Industry
  • Data Warehousing
    • Oracle, IBM
    • Community Edition - Infobright
  • Software Agents
    • Still in research but consider Google and Wolfram|Alpha
  • Statistics and Data Mining
    • SPSS, SigmaPlot (the big boys)
    • Open Source - R (+ UI, e.g. R Commander)
  • Business Intelligence (BI)
    • IBM TM1, Cognos, Tableau
    • Open Source Pentaho, and Palo
  • Other
    • Matlab, Matematica
    • Excel

Data Visualisation

• One aspect
• Edward Tufte - The Visual Display of Quantitative Information.
• Junkcharts - A blog about good chart design.
• Chart Porn - A blog about (content-light, but pretty) infographics.

Aside: Open Data

• Open APIs
• Programmable Web (Google Maps, twitter, Flickr).
• Open government data initiatives (e.g. NZ).
• Limited access to raw data/analysis code from Scientists. Call for a change

• Using Open Data
  • IBM Many Eyes
  • Swivel

The Crux: Data Visualisation

• Basic Libraries
  • Processing (and processing.js).
  • The venerable gnuplot
  • Graphviz
  • Many others... 25 for the web
  • My current favorite: Protovis (examples)
• Data Analysis
  • Sage (matlab alternative. Python)
  • R (statistical analysis)
  • Prettygraph

R-Node

• A web frontend to R. (http://www.squirelove.net/r-node).
• Why?
  • Because I needed something to do in the evening after my daughter went to bed.
• Really, why?
  • Because I wanted a way to leverage the power of R in their web-based systems.
• demo
• Future
  • There is a need for R to be accessible as a service to users
  • I will focus on R-Node becomming a strong web-based UI for a central R service.

Thanks

• Try out R-Node at http://squirelove.net:2904/
• Join the discussion list at http://gitorious.org/r-node
• Get the code at http://groups.google.com/group/r-node-users
• Visit the website at http://squirelove.net/r-node/