Week 10: Accounting Automation

Two threads this week. The lecture covers clustering, the last new method of the term, then bridges into accounting automation and what it means for a process to be deterministic. The graded work is the automation itself. You build a revenue journal entry from about a hundred daily transaction files, screen the odd rows, and let the total recompute.

Lab 10 and Homework 10 are both due this week, and Project 3, the capstone, opens. Check Canvas for the due dates.

Lecture

• Accounting Automation: Clustering and the Deterministic Line (interactive lecture)
  Self-paced deck. Clustering from the ground up, how it groups transactions without labels, then the bridge into accounting automation and where a process is deterministic and where it is not. Every slide carries an in-deck AI tutor and a self-check.
• Recorded lecture (YouTube)
  The webinar recording. Watch it if you missed the live session or want to walk back through the deck at your own pace.

Lab

• Lab 10 — Automate a Journal Entry
  The walkthrough for the vibe-code lab. Build a revenue journal entry from the daily transaction files, flag the outliers that do not belong, and keep the odd rows that do. The journal entry total is checked against a fixed answer, so the number has to land.

Lab dataset

• transactions.zip
  The transaction files for the lab, zipped. Download and unzip, or open the starter below straight in Colab and skip the download.
• Colab starter notebook
  Open the starter in Google Colab. No install. It reads the transaction files and gives you a place to build the entry.
• Data dictionary
  What each column in the transaction files means. Read it before you start screening.

Practice

• Transaction Review (SQLite workbench)
  Query the lab transaction data in the browser with SQL. No setup. A fast way to eyeball the data and find the rows that look wrong before you write any code. Carries the AI tutor.

Optional references

• Guiding Gallery
  Worked examples of cluster visualizations. The charts that read clearly sit next to the ones that do not, so you can see what separates them.
• Unstick Yourself
  A debugging page for when the notebook throws an error you do not recognize. Common failures and how to read them.
• Power BI reference
  An optional reference on Power BI, if you want to see this kind of work in a dashboard tool.
• The Recompute Stretch
  An optional build. Recompute a model by hand, one rung at a time, then wire it into a dashboard. This is the lineage of an earlier version of the week, kept for anyone who wants to go further.

This Week