I used a the FAMD algorithm of R on my own dataset.
My dataset contains about 10 columns of numerical(float) data, 6 columns of one-hot-encoded data and 1 column with categorical data (zip codes) with a lot of unique string values(unique values>8000). I use the following code:
res.famd <- FAMD(Data, ncp=5, graph = FALSE) print(fviz_screeplot(res.famd,addlabels=TRUE)) print(fviz_famd_var(res.famd,labelsize = 2, repel = TRUE)) print(fviz_contrib(res.famd, "var", axes = 1)) print(fviz_contrib(res.famd, "var", axes = 2)) print(fviz_famd_var(res.famd, "quanti.var", col.var = "contrib", gradient.cols = c("#00AFBB", "#E7B800", "#FC4E07"), repel = TRUE))
This is pretty much their example from the tutorial website and it works fine, however my results are far from understandable.
The problem is, when I print the scree plot, the cumulated percentage of explained variance is less than 0.5% for all features combined. I tried to increase component number, but this did not produce any changes in my results.
I do not know the reason for this, but I suspect that the categorical zip codes, which are probably converted into some kind of scaled one-hot-encoded columns, distort the result?
Maybe you can give me a hint, since I do not find much information on how to interprete these results - or maybe I just made a mistake in my code/dataset and you can help me out.