# Predict multiple labels when labels are of mixed type: floating point and categorical

I was provided with a data file which looks like this:

f1,f2,f3,f4,lbl1,lbl2,lbl3
-99.78507814566231,0.9487271429068756,6.354574478964145,-0.92299976023887444,-1.0,7.785451430712673,category2
-100.45507785716314,2.780304246171124,3,131265316135844,2.43832325135605543,-1.0,5.229512156937786,category3
-80.35411219282111,3.31425742741509244,4.12266026421345,15.422132963016145,1.0,9.45646860327132,category1
[...]


There are four columns of features and the last three columns are labels that I need to predict.

The exercise involves the following questions:

1. We suspect that there's correlation between the 3 labels. Investigate that correlation.
2. Can the task be solved by making only one algorithm that predicts one of the label values, after which the other labels can be deduced?
3. If so, which label should be predicted by an algorithm and which can be deduced from it?
4. What type of supervised learning would be suitable for predicting the selected label column and why?

I feel kinda lost because labels seems to be of mixed type: two of them are floating point values, the last one - categorical.

• How can I look into correlation,
• How do I pick just one label
• What supervised learning would be suitable to make predictions here?