ASK HN: Predicting Time Series Data

4 points by redlion 4y ago ↗ HN

I want to forecast traffic on my website (i work for a large company with thousands of hours spent on our website hourly across the USA). I am researching how to visualize and predict this time series at the hourly level. Any ideas here on modeling the time spent on the website on a city level? I have lots of historical data ~2 years

5 comments

[ 4.0 ms ] story [ 15.9 ms ] thread

temp234 4y ago ↗

Sounds similar to problems predicting hourly call volume. Statistics research might give you examples, also operations management research

uaas 4y ago ↗

After you have the proper aggregations, you can use stats methods e.g.: https://about.gitlab.com/blog/2019/07/23/anomaly-detection-u...

unlikelymordant 4y ago ↗

https://facebook.github.io/prophet/

Prophet is a pretty popular library for time series forecasting

Jugurtha 4y ago ↗

We have that as an application on our platform. We did that because we use it often and we've included it as a plug-in. We made it addressable by API because sometimes we hook it to a Grafana dashboard and use it in other systems, but you can also upload a file (CSV, spreadsheet) and get predictions and graphs.

Feel free to contact me.

methusala8 4y ago ↗

This book here suggests using Dynamic regression :

https://otexts.com/fpp2/weekly.html