search

ruiking04 / COCA

Deep Contrastive One-Class Time Series Anomaly Detection
30 stars 9 forks source link

COCA Visualization Plots #26

Open marciahon29 opened 9 months ago

marciahon29 commented 9 months ago

The plots that I do by hand look like this: Electricity Consumption - 000 - Original

The plot generated by COCA looks like this: IESOToronto202120222023_9_Image_00

The plots are not the same. In fact, the COCA is much smaller than my plot.

Please could you let me know what I should do to make the visualizations to work properly.

marciahon29 commented 8 months ago

Hello, I think I understand. Your plot only covers the last 5000 points.

ruiking04 commented 8 months ago

Hello, I don't see the anomaly score in your graph. I would suggest another y-axis to represent the anomaly score.

Screenshots and links to the code are as follows：

image

code

marciahon29 commented 8 months ago

The graph generated by your code does have an anomaly axis. The graph that I generated in python does not.

I was showing the two graphs in order to show the similarity between the two.

It's okay, I think I understand the problem. Your graph only covers the last 5000 points whereas my graph covers the entire timespan.

marciahon29 commented 8 months ago

Additionally, in the second plot generated by COCA code, you notice that the anomaly score is either a "0" or a "1". Does this sound correct and possible? I look at your image in the paper and it is more spread out.