I'm trying the ML beta on a sample of the kind of data I wish to analyse and track with it. I'm not sure if I've got it set up correctly but I can't see this working properly for the traffic profile.
I can see how the blue shaded area starts to track the values at the start (which is a weekend), but when we get into the working week, the traffic pattern seems to make the model wildly inaccurate. For example, part way through the week, the model start to accept -100,00 as an acceptable value. I'm hoping we can get this working to track the traffic line very closely, and take into account the drop in traffic levels over the weekend.
Hi Steve. The model needs to see 2-3 periods before it detects a periodicity type. For daily periodicity, that means 2-3 days. For weekly periodicity, 2-3 weeks. In your case, we're talking about weekly periodicity (on top of daily periodicity). Could you please feed 3 (or more) weeks of data and see what happens?
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.