StackExchange Tag Recommendation

Members: Dingxian, Yida, Vy, Peiyang, Kirthanaa
Dec. 12, 2016

Background & Goals

  • Post direction process: Tags
  • Hypothesis: Good tags can attract good answers
  • Tag recommendation: Manual assignment Vs Automatic assignment

Data & Cleaning

Methodology

Hypothesis Testing

  • \( H_0 \): Good tags can attract good answers
  • Under \( H_0 \), build Tag recommendation system by KNN method as good tag producer
  • Regression tag score(wacc) with respect to answers score(Score_a)
  • Rejection rules: if the coefficient in the simple regression is not significant or positive then we should reject the null hypothesis.

Result: Recommendation system

Result: Recommendation system

Result: Score_a ~ wacc

Result: Hypothesis testing

Good tags can attract good answers!

Limitations

  1. Simple regressioin too simple
  2. Information wasted for simple recommendation system