Modelling Experts Behaviour in Q&A Communities to Predict Worthy Discussion

Summary

This project is part of a scientific research. We are interested in knowing the influence of experts in discussions. You can find here the analysis of the paper "Modelling Experts Behaviour in Q&A Communities to Predict Worthy Discussions" published at the 17th IEEE International Conference on Advanced Learning Technologies - ICALT 2017. In other words, you will find here everything you need to reproduce this research. First, read the paper and then follow the instructions below.

Environment

• R version 3.2.2 (available at https://www.r-project.org/)
• MySQL database version 5.6.31 (available at http://dev.mysql.com/downloads/)
• Ubuntu version 15.10
• HeidiSQL version 9.3.0.4984

Inserting data into MySQL

• Open folder database/biology
• Unzip bio-database.sql.zip
• Run bio-database.sql

How did we extract the data?

We used the stackexchange API (http://api.stackexchange.com/docs/) to extract data from biology Q&A (http://biology.stackexchange.com/). Then, we loaded the data into MySQL.

Analysis

For the analysis, each one has a folder with the SQL that generated the CSV file. We used the CSV as the input for our R script.

A. Reputation and Participation Analysis

• Open folder A
• Execute: Rscript reputation.R

B. Quality of Interactions

• Open folder B
• Execute: Rscript qualityInteraction.R

C. Users Interactions and Evolution Analysis

• Open folder C
• Execute: Rscript evolution.R

D. Best Answers Analysis

• Open folder D
• Execute: Rscript best.R

E. Discussion Length Analysis

• Open folder E/part-1
• Execute: Rscript discussion.R
• Open folder E/part-2
• Execute: Rscript discussion.R

F. Graph and Correlation Analysis

• Open folder F
• Execute: Rscript graph.R

G. PREDICTIVE MODEL TO FINDING A WORTHY DISCUSSION

• Open folder G/nnet
• Execute: Rscript prediction-gbm.R
• Open folder G/gbm
• Execute: Rscript prediction-nnet.R