Approach - sorrachai/FraudResumeDetection GitHub Wiki

After preprocessing,

• Resume is a set of sections;
• Sections contain a bag of key words

Data Mining Tasks (Training sets)

• baseline: search for k-nearest pairs using jaccard similarity or q-gram.
• our approach: (int t)
  1. search for t-nearest pairs using same as above
  2. among t candidates, apply Naive Bayes' classifier using indicators
• indicators:
  1. (different owner*) plagiarism between sections (Using graph maximum matching, parameter: threshold)
  2. university name (parameter: list of faked university)
  3. DOB vs. claimed experience (parameter: threshold)
  4. (same owner*) Inverse of plagiarism between sections (parameter: threshold)