How can I classify columns of a table

4 views (last 30 days)
Sara on 8 May 2018
Answered: Bernhard Suhm on 30 May 2018
I have a csv file with several columns ( current, voltage, time, temperature). I want to classify each column separately, I mean I want to consider each column as a class so that in case of new data it is automatically labeled as current, or voltage, or time... . Is it possible with machine learning techniques? please suggest me a suitable function.

Answers (1)

Bernhard Suhm
Bernhard Suhm on 30 May 2018
You could try to train a classifier by considering each data point of your table an example for the corresponding column's name (class), you'd have four classes current, voltage, time, temperature. But that's not going to be a very strong predictor, just to the extent the actual values are in disjoint numeric ranges will the classifier be able to distinguish them. To the degree there's overlap, the classifier will just guess randomly if a new value falls within the overlapping range. It doesn't really matter which classifier you use, since this is fundamentally not a good use case. For a better predictor you need to feed it more hints, like is there any patterns in how you receive new data points, a certain number of currents followed by a different number of voltages etc?

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by