https://ndjstn.github.io/I build practical systems across software, automation, local AI tooling, data workflows, and technical problem-solving. 2026-04-24T04:03:24-05:00 Justin Stone https://ndjstn.github.io/ Jekyll © 2026 Justin Stone /assets/img/favicons/favicon.ico /assets/img/favicons/favicon-96x96.png 2026-04-24T00:00:09-05:00 2026-04-24T03:38:22-05:00 https://ndjstn.github.io/posts/youtube-top995-five-archetypes/ Justin Stone

KMeans on the Global YouTube Statistics 2023 dataset produces five distinct creator archetypes: mega-scale, mainstream, low-engagement, music-video, and upload machines.

2026-04-24T00:00:08-05:00 2026-04-24T04:02:59-05:00 https://ndjstn.github.io/posts/olist-ecommerce-retention/ Justin Stone

Olist's 2016-2018 marketplace data shows 97 percent single-purchase customers. Revenue growth is all acquisition; retention is near-zero by construction.

2026-04-24T00:00:07-05:00 2026-04-24T03:38:22-05:00 https://ndjstn.github.io/posts/bike-sharing-timeseries-cv/ Justin Stone

Random 5-fold CV makes LightGBM look like the winner on UCI hourly bike-sharing data. Time-series CV says a naive seasonal-mean baseline beats it.

2026-04-24T00:00:06-05:00 2026-04-24T03:38:22-05:00 https://ndjstn.github.io/posts/nyc-taxi-geography-beats-odometer/ Justin Stone

Distance alone predicts 64 percent of trip duration. Adding geography and time of day gets you to 80 percent. The heatmap shows where the missing signal lives.

2026-04-24T00:00:05-05:00 2026-04-24T02:34:45-05:00 https://ndjstn.github.io/posts/house-prices-ames-stacked-ensembles/ Justin Stone

A stacked Ridge + XGBoost + LightGBM ensemble on Ames houses reaches 0.126 RMSLE, beating a Ridge baseline by about 8 percent. The stack weights are unbalanced, and that's fine.