Hello!
I've been trying to run anomaly detection on the data I have, but I've been strugling with some anomaly likelihoods results. I got the parameters for the algorithm using getScalarMetricWithTimeOfDayAnomalyParams. Firstly, here my data plotted. The x axis is the time, in milliseconds, and the y axis the scalar values:
Here the x axis is the time, in milliseconds, and the y axis raw anomaly score:
And here the x axis is the time, in milliseconds, and the y axis the Anomaly Likelihood:
So, i thought those result might erroneous, than I inserted some error between 7001500-7001600 data point, here follows the data plotted, raw anomaly score and Anomaly Likelihood in this sequence:
So when I inserted the error I increadsed the signal within 100 points in range of 7001500 to 7001600 for 2.000 units, but the spike only reached 0.2 in Anomaly Logscore units. This behavior is right (I was expecting a higher anomaly likelihood in the interval where the error was inserted) ? I think I probably made a mistake on my Anomaly parameters. Can someone point to me where and what I should look for mistakes in my code? Here are the Anomaly parameters i used:
anomaly_score = Anomaly(
slidingWindowSize = 10)
anomaly_likelihood = AnomalyLikelihood(
learningPeriod=500,
historicWindowSize = 8640)