0
$\begingroup$

I've got a data set that looks like this

physical_data1   physical_data2  switch1   switch2   state
400              500             1         0         Normal
400              500             1         1         Normal
500              650             0         0         Normal
600              700             1         0         Normal
1000             300             1         1         Anomaly!

where physical_data are data that are from 0 to 1000 and switch are switches that are binary (1 means on 0 means off).

I am still relatively new to machine-learning and so I am wondering what kind of machine-learning algorithm is best suited for these kinds of data to detect anomalies, where my data has mixed features of physical and binary quantities.

Things that i have done are, normalizing to [0,1], however i am not quite sure if applying PCA to this kind of data will result in a loss of detection rate because it requires all features to determine if a particular reading is normal or an anomaly.

One other question I have is that what if in my data set i do not have the Anomaly data, instead, only have normal data. In that case what kind of model can i use?

$\endgroup$
0
$\begingroup$

Try Naive bayes. It's simple and can be used, if the features are independent of each others.

$\endgroup$
0
$\begingroup$

Any kind of classification algorithm should be alright for your data.

Logistic Regression is probably the first thing you will learn in a ML tutorial.

https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.