Skip to content

Project Instructions

WEDNESDAY: Complete Workflow Phase 1-3

Follow the instructions in ⭐ Workflow: Apply Example.

Complete:

  1. Phase 1. Start & Run – copy the project and confirm it runs
  2. Phase 2. Change Authorship – update the project to your name and GitHub account
  3. Phase 3. Read & Understand – review the project structure and code

FRIDAY/SUNDAY: Complete Workflow Phases 4-5

Complete:

  1. Phase 4. Make a Technical Modification
  2. Phase 5. Apply the Skills to a New Problem

Topic

Drift detection for monitoring changes in system behavior.

In this project, you compare recent system behavior with a baseline reference period to determine whether the system has changed.

Drift detection helps analysts determine when a system is behaving differently than expected.

Learning Objectives

After completing this project, you should be able to:

  • Explain what drift means in a monitoring system
  • Compare current observations to a reference baseline
  • Detect when system metrics shift away from normal behavior
  • Run and validate a professional Python project
  • Interpret evidence that suggests system behavior has changed

Example Code

The example file is located in:

src/cintel/drift_detector_case.py

It demonstrates:

  • reading a reference dataset representing baseline behavior
  • reading a current dataset representing recent observations
  • computing summary statistics for both periods
  • comparing the two periods to detect drift
  • writing results to an artifacts file
  • logging the pipeline process

Run the example and review the code before creating your own version.

Dataset

The example datasets are located in the data/ folder.

Typical files include:

  • reference_metrics_case.csv
  • current_metrics_case.csv

The reference dataset represents normal system behavior.

The current dataset represents more recent system activity.

Drift detection compares these two periods to determine whether the system has changed.

Your Phase 4: Technical Modification Task

Using the example as a guide:

  1. Copy src/cintel/drift_detector_case.py.
  2. Rename the copy to src/cintel/drift_detector_yourname.py.
  3. Run your copied file to confirm it executes correctly.
  4. Modify the program by changing how drift is evaluated.

Possible modifications include:

  • adjusting the drift threshold
  • comparing additional metrics
  • changing the summary statistic used for comparison
  • adding additional logging or output columns

Then:

  • run the project
  • confirm the drift results appear in the output artifact
  • confirm the program logs useful messages

The goal of this phase is to practice modifying a working drift detection pipeline.

Phase 5: Apply the Skills

In Phase 5 you apply drift detection to a new situation.

Possible approaches include:

  • comparing two different time periods in a dataset
  • detecting drift in website traffic or system performance
  • evaluating whether error rates or latency have shifted
  • monitoring changes in operational metrics

Update your documentation in docs/ to explain:

  • what datasets you compared
  • what drift indicators you used
  • what changes you observed
  • what those changes might mean for the system

Drift detection is an important component of continuous intelligence, helping analysts identify when system behavior begins to change.

If you would like to apply these skills to a real dataset instead of the provided example data, see suggested datasets:

https://denisecase.github.io/pro-analytics-02/reference/datasets/cintel/