Anomaly Window Detection

Problem

Count sliding windows of length window whose mean deviates from the overall mean by more than k population standard deviations: abs(window_mean - mean) > k * std.

Return 0 if window is not positive, window exceeds the series length, or the series has zero variance.

Hint

Compare each window mean to the global scale, and use a strict inequality.

Test cases

case 1

in: [[1,1,1,1,10,10],1,2]

expected: 1

case 2

in: [[1,2,3,4,5],0.5,1]

expected: 4

case 3

in: [[5,5,5,5],1,2]

expected: 0

case 4

in: [[1,2,1,2],10,2]

expected: 0

Starter codePython

def anomaly_window_detection(series, k, window):
    # Your code here
    pass

4 test cases · run your code against them in the editor