It's odd but most of my datafantic newsletter subscribers are in Europe, but Europe only accounts for 30% of traffic.
Maybe I'm less of an American than I used to be.
| datafantic | https://datafantic.com |
Robert Ritz
- 211 Followers
- 269 Following
- 61 Posts
Data scientist and educator based in Mongolia. I share my projects on datafantic.com.
I am so potato at marketing.
I posted my Airbnb scraper on Reddit and literally thousands of people viewed by code.
I had a link to the blog post the whole time, but just added a link to signup directly in the notebook yesterday. Result?
Newsletter signups!
Make it easy for people to sign up!
It’s great when your kid is growing up. They can sit at the table and do their homework on their own while I unwind from the day.
Then we can play Minecraft dungeons together.
Simon PhippsSeeing people with coding ability commenting on features they want in Mastodon, many of them framing their insights as if there is a bug.
Almost all ideas I have seen so far could be implemented in a client or a relay. For ones that can't, there are pull requests. For ones that are declined, downstream forking is feasible.
It's #OpenSource & based on a truly open standard, #ActivityPub. How about we frame ideas positively & look for places to contribute? This is not a walled garden run by VCs.
I'll still publish my projects at datafantic.com and share all the code and data as usual.
Just no live streaming. Future project write-ups will focus more on:
- How the project was done,
- How I thought through it (and what changed)
- What could have gone better
- Future possibilities for the project
After trying to live stream my real estate forecasting project last week, I've decided to scrap the idea. A few reasons:
1) I don't think clearly on the live stream. This project took a lot of original thinking.
2) Data projects take a lot of time. Each project will be 5-8 hours of video on YouTube. I wouldn't watch that.
3) Finding a quiet space for an hour per day when I'm a university administrator and instructor is super hard.
4) I like listening to music while I work.
Data from the image is from the hold out test set.
My biggest question: Is this even useful? My thoughts:
- Widening confidence interval is a signal of volatility, but this is obvious from the inherent volatility of the time series.
- Predictions are better than random but fail to predict spikes.
- Predictions like these could potentially influence the market, creating self-fulfilling prophecy or the opposite.
Want to see the code?
I'm working on a hobby project to predict the median sale price of homes in cities around the US 3-months ahead.
Post-COVID, the median sale price changed an average of 8% every 3 months.
My current best model gives an average % error of 6%.
The picture below shows Plano, TX. Blue lines are average predictions + 80% CI. The red line is actual values.
I'm not super happy with this, but it's not too bad.
🧢📚🚵♂️⚙📈☕Last week's #TidyTuesday looked at radio stations around the United States and their reach. Code can be found here: https://opus1993.github.io/myTidyTuesday/. Credit to Frank Hull for the dataset and @kylewalker for {tigris}. #rstats #r4ds
Todays stream revealed I can predict price per square foot of US homes 3 months in advance post COVID with an avg error of 11%.
So basically I learned that I can’t predict home prices.
You’re welcome.
Pre COVID model error is 5%. So sorta good.
Question is, does price prediction actually help? If you knew prices were trending up or down would that change anything?
