CODEWITHSUNDEEP
machine-learningneural-networkartificial-intelligenceclassification
AdeB
AdeB

Reputation: 187

Classifier with heterogeneous data

I have a l2-dimensional data set of 1000 samples composed of 5 temperature values, 5 price values, one integer value representing a judgement by a human expert (undecided=0, good=1, bad=2, danger=4) and a binary decision variable that I want to learn to predict.

How can I find a classifier than can cope with this heterogeneous data ?

I was thinking about building one classifier for each possible human judgement (0,1,2,4), so 4 classifiers. So for each human judgement value, I would: - center and reduce the temperature and price values - maybe use PCA to remove some irrelevant features - use a machine learning method for classification (like multi layers neural networks or SVM)

Is my approach correct ? (what if there were 1000 possible human judgements instead of 4 ?)

Upvotes: 1

Views: 959

Answers (1)

alfa
alfa

Reputation: 3098

A typical way of encoding categories for SVMs or ANNs is the 1-of-C encoding:

Generally almost every classifier can deal with heterogeneous data. But you have to preprocess the inputs (scale, normalize, ...). There should be plenty of hints in the links I gave you.

Upvotes: 2

Related Questions