Search code examples
pandasbooleanone-hot-encoding

Compare one-hot-encoded column header and predicted labels

I have 3 one-hot-encoded columns where the header names are labels, and one prediction column preds where the labels are predicted (see image). I want to calculate the performance of my predictions by comparing the label in preds and the 1-encoded column header.

see here

In this example I only have 20% predicted correctly.

Is there a quick way of doing this in pandas?

Solution

  • IIUC, DataFrame.lookup and np.mean

    df[['Type_1','Type_2','Type_3']].lookup(df.index, df['preds']).mean() * 100