Dataset Overview

Dataset File Structure

• Training set files:

  • train_ev.txt: Types of actions performed by the user.
  • train_ti.txt: Timestamps of the user actions.
  • train_go.txt: Final goals of the activities.

• Testing set files:

  • test_ev.txt: Types of actions performed by the user.
  • test_ti.txt: Timestamps of the user actions.
  • test_go.txt: Final goals of the activities.

Data Processing

• Preprocessing: Utilizes the develop_dumps.py script to merge the separate files, normalize action timestamps, and generate .p format data for training and testing.

Model Evaluation Metrics

• Accuracy (Acc): Accuracy of predicted event types.
• Mean Absolute Error (MAE): Average absolute error between true and predicted action times.
• Goal Prediction Accuracy (GPA): Accuracy of goal prediction on the test set.
• Interval Goal Prediction Accuracy (Itv. GPA): Reflects PROACTIVE's ability to correctly predict the goal at each new action arrival, distinct from overall goal prediction accuracy, used to track performance changes caused by the gamma (RL‑trick).