Logistical regression for system of coordinates

Question

I am learning to apply different machine learning techniques. Now, I was recommended logistical regression for my problem, but I can't figure out how to apply it to my dataset.

I have a dataset with xyz-coordinates as a time-series, where each row contains the xyz data for one test-subject, labeled with either A or B. I would like to classify whether a new row of xyz-data would belong to condition A or condition B. Is it possible to apply logistical regression to my data, when I have to classify by looking at a time-series based set of coordinates? E.g. can I use logistical regression to predict if a set of time-series coordinates belongs to condition A or condition B?

The following array is an example, where each set of three coordinates represent the XYZ-coordinates of that time frame.

array([array([[array([[0.01889996]]), array([[-0.28033654]]),
    array([[1.4816033]]), array([[2]], dtype=uint8)],
   [array([[0.01381396]]), array([[-0.28270772]]),
    array([[1.50570126]]), array([[2]], dtype=uint8)]

Answer 1

You can use feature engineering concept known as lag-feature, to keep sequence relation. For example, the label for coordinates will probably depend on the last or 2nd last coordinates. So you can use 2 last rows of features along with current row, with output of current row as target. This is an efficient method to keep such co-relation in data. Here is link for more info.

You can also try recurrent-network, which is another best way to keep track of time-relation in data.