3 ms·

Show HN: Extended Isolation Forest for anomaly detection

• Macuyiko 8y ago
  Very nice. Does this also solve the original drawback of IF of not being able to handle categorical values? You could create dummy vars but this would lead to somewhat biased results using the standard orthogonal splits. Would EIF work better here?
  • mgckind 8y ago
    Thanks! We were concerned about continuous variables. For categorical values there might be some improvement although it might require extra attention not to be bias due to the dynamic range or cardinality of the values.
    • Macuyiko 8y ago
      That makes sense, thanks!