Skip to content
Yun edited this page Jun 18, 2015 · 23 revisions

OUTPUT FILE

  • "XXX_Prediction.csv"
    Currently, it is ordered by KC, student (and preserving the record order within a KC-student), which may not be the same as the original ordering in test set (unless you just have one HMM in test set). The columns actualLabel,predLabel,predProb,probKnow,student,kc are:

    • actual student responses(correct or incorrect).
    • predicted student responses.
    • predicted probability of getting correct responses.
    • predicted probability of known/mastered before updating by current observed correctness.
    • predicted probability of known/mastered after updating by current observed correctness (before doing transition to next hidden state).
    • student
    • kc
    • problem
    • step
  • "XXX_Evaluation.csv": Evaluation files including:

    • predictive performance metrics on test set
    • plausibility metrics (see our EDM15 paper mentioned in related work in this wiki).
    • training information.
  • "XXX_Parameters.csv": This file maintains the HMM basic parameters and feature coefficients.

  • Runtime.log: This file contains the basic runtime information (configuration, log likelihood changes, etc.).

Clone this wiki locally