python 监督学习与数据集分类

Question

I am new to deep learning and am currently researching a certain topic. I am looking for machine learning detection of anomalies in time series pattern and their implementation in python.

For example, I have a recording of the different CPU frequencies of my computer during a certain time interval. I would like to implement a supervised learning algorithm that takes a time series of CPU frequency as an input and decides, whether anything "unusual" happened during that time (unusual CPU usage etc).

EDIT:

My data sets look the following way, every 10 seconds the current CPU frequency is measured. I have not specified an exact number of datapoints per set, the following is just for illustration. But I am expecting around 2500 datapoints per set:

Dataset_1: {1.2, 1.2, 1.6, 1.3, 1.5, 1.7, 1.6, 1.4, 1.5} -> Label: "good"

Dataset_2: {1.3, 1.2, 1.4, 1.3, 1.4, 1.5, 1.9, 2.1, 2.0} -> Label: "good"

Dataset_n: {1.3, 1.2, 3.6, 3.5, 1.4, 1.5, 3.3, 3.2, 1.2} -> Label: "bad"

My understanding of a supervised ML algorithm is that i have training datasets. However, every tutorial that i have found so far always labels each value in a data set. In my case that would not be possible, as I could only tell my ML algorithm:

a) this time series data set is normal

b) in this data set something is not normal

but i wouldn't be able to label each individual value, meaning i cannot say:

1.2 -> good

1.3 -> bad

1.4 -> good

As there are many different ML algorithm, it is hard for a beginner to determine which is a good one to use. So my question is:

Which (python implemented) algorithm could i use as a start, that accepts labels for entire datasets and does not expect each value to be labeled.

I hope this question makes sense, edits are highly welcome as much as your time! thanks!

Answer 1

For this application I would go with KNN(K - nearest neighbors). Tech with Tim has a great tutorial on KNN, explains it well and shows the implementation. Hope this helps