Thread by @DataChaz

Alrighty then! Here's StreamProphet in über early Beta! 🤗

- Connect to GSC
- Visualise forecasted #SEO traffic + time series' trends & seasonality
- Export predictions to CSV + models to JSON! 🔥

🎲 App: bit.ly/3a6oliw
📬Post: bit.ly/3sfSoKT

@streamlit

🧵↓

https://twitter.com/DataChaz/status/1379428389262557188

🚀 How to get started!

- Click the 'sign-in w/ Google' button to reach the consent screen
- Copy the Oauth token & paste it back in the app
- Type the web property you want to review
- Define your forecast horizon
- Press the 'Fetch GSC data & get predictions' button! 🔥

2/8 ↓

📈 Checking predictions

- You can visualise past data & predictions in the right-hand side chart
- Left charts show the decomposed trend, weekly & yearly seasonality (when avail.) of the time series
- You can export predictions via CSV + your forecast model via JSON 🤗

3/8 ↓

🛠️ Here’s the stack I used:

- Web framework: the mighty @streamlit! 🎈
- Forecasting library: @fb_engineering Prophet
- PyStan
- Pandas
- 100% Python! 🐍🔥

4/8 ↓

🌅 Early Beta

- StreamProphet's still in early beta, with possible rough edges! You. Have. Been. Warned! 😊
- Issues, feedback or suggestions are welcome. Head-off to my Gitter page for a chat: gitter.im/DataChaz/Strea…

5/8 ↓

📝 To-Dos

- Add session state so parameters can be tweaked w.o. having to re-authenticate
- Add seasonality params.
- Holidays import so they can be factored in the forecast
- Add cross-validation to optimise hyperparams
- More forecasting models: ARIMA, HWAAS & more!🔥

6/8 ↓

🙌 Shout-outs!

Kudos to @JoshCarty and @_okld for their precious help!

Thanks also to @jakubmotyka, @KorayGubur, @periklispapanik and @RichardFergie for that early testing! 😊

7/8 ↓

Free as in beer!

Lastly, this app is free. If it's useful to you, you can buy me a 🍻 to support my work! 🙏

👉 buymeacoffee.com/cwar05

8/8

@threadreaderapp unroll