Clojure Emerging Data Ecosystem
summary
A Clojure REPL workflow loads 18,000 messages (~250 users) from Zulip and uses tablecloth + tablecloth-time for fast columnar feature engineering (datetime, next-response-time, activeness, sentiment) and Hanami/Vega for plots. A linear regression on log next-response-time gives modest correlation (R≈0.5 on the two variables visualized but R^2≈0.28 on logs, negative on raw seconds), while a cycloheml logistic regression classifies “active” threads at roughly 73% accuracy; code and pipelines run end-to-end.
AI-written summary · about
chapters (12)
- 0:00Start
- 9:19Timestamp to Localdatetime
- 10:10Datetime Conversion
- 11:12Emap
- 14:09Next Response Time
- 17:20Format the Data
- 17:34Plot Type
- 19:01Adjust Frequency
- 19:56Line Chart
- 26:58Linear Regression
- 33:51A Classification Pipeline
- 35:00Confusion Matrix
related talks
44mData Science in Clojure
43mDoing Data Science with Clojure
31mClojure at 4,000 msg/s-What We Learned, Loved, and Loathed
26mDoing data science with Clojure: the ugly, the sad, the joyful
24mThe Return of Clojure Data Science
37mDon't Shed a Tier for Me, Clojure
28mEvery Clojure Talk Ever
38mScientific Clojure, a bird's eye view