check_pattern() in the sequences.py

[echozheng]

Hello again! I am interested in the SEE algorithm, and reading your code. And here i got some doubts about it. It's the code in the function check_pattern(), sequences.py.

1: if sequence_class == multivariate_timeseries.class_timeseries: 2: #tp, fp, tn, fn, time_steps 3: if pattern_found: 4: return 1, 0, 0, 0, time_steps 5: return 0, 0, 0, 1, [] 6: else: 7: if pattern_found: 8: return 0, 1, 0, 0, time_steps 9: return 0, 0, 1, 0, []

In the 5th line, while we get the right classification of multivariate_timeseries ，why the function returns False Negative？ Suppose a multivariate timeseries has the list of shapelets as ：s1，s2, s3。and both shapelets matching index is 1. Of course the pattern_found is false, but isn't it still a sequence which could do the right classification only without an order between shapelets.

Thanks for your time.

[rmgitting]

Hello, Thanks for your interest. I may not completely understood your question, but I will try to clear it. If the time series has the same class of the pattern, but the pattern is not found in this time series, then from the perspective of this pattern it is a false negative. That is, it is not detected by the current pattern, even though they have the same class. But the order is already taken care of before that, for example in the code: indices = [index for index in aShapelet.matching_indices[multivariate_timeseries.name] if index >= match_index]

Hope I answered your question.

[echozheng]

Yeah, I get the answer just I wanted. Could you please describe some details about the database used in the project? What is wafer and what the every dimension means? Thanks.