Error when run DenStream.runOnNewSample() without run DenStream.runInitialization()

[luisalfarob]

Hello,
When I try to run your code without runInitialization function I get the follow error:
Traceback (most recent call last):
File "C:\Anaconda3\lib\site-packages\IPython\core\interactiveshell.py", line 2881, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "", line 5, in
result.append(den1.runOnNewSample(sample))
File "C:\git\OutlierDenStream\DenStream.py", line 178, in runOnNewSample
self.buffer.append(sample)

This error can be resolved changing this line:
self.buffer.append(sample)
for:
self.buffer.append(sample.value)

Solved!
Now I get another error:
Traceback (most recent call last):
File "C:\Anaconda3\lib\site-packages\IPython\core\interactiveshell.py", line 2881, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "", line 5, in
result.append(den1.runOnNewSample(sample))
File "C:\git\OutlierDenStream\DenStream.py", line 179, in runOnNewSample
if (len(self.buffer) >= self.numberInitialSamples):
TypeError: '>=' not supported between instances of 'int' and 'NoneType'

I noted that the variable numberInitialSamples is not defined. So I'm thinking if I need change the way that a run the code or there are bugs when try to run the code for multicluster operations...

I hope you can help me with this problem.
Regards.

[anrputina]

Hello,
Unfortunately, the algorithm has a very specific purpose right now (anomaly detection) so the initial clustering part is not very well developed. In particular, for now, the algorithm manages two clusters: Normal or abnormal. So if it is able to merge the new sample to the core-micro-clusters the output is 'Normal', 'Abnormal' otherwise.

In the next updates I can add the multicluster support (now, unfortunately, I work on something else). If you wanna contribute any help is welcome. If you try to use the multicluster operations now I think you're going to obtain only bad results.

For what concerns your error: self.numberInitialSamples is initialized to None thus you have to assign a value during the initialization phase, something like this:
den = DenStream(lamb=0.03, epsilon='auto', beta=0.03, mu='auto', ..., numberInitialSamples=XX)
Hope this can help you, but anyway the code is not ready for multiclustering (just 2 clusters: normal and abnormal in my case.)

Write to me if you need anything else.

[ZahoorAhmad]

Kindly let me know how to run for anomaly detection there is no help for startingBuffer in initialization..
Thnx in advance