Building Machine Learning Systems with Python - Third Edition

Training the classifier

Let's use the logistic regression classifier, which has already served us well in the Chapter 9, Classification II – Sentiment Analysis:

from sklearn.linear_model.logistic import LogisticRegressiondef create_model():    return LogisticRegression()

Just to mention one surprising aspect: the evaluation of accuracy rates when first switching from binary to multiclass classification. In binary classification problems, we have learned that an accuracy of 50% is the worst case, as it could have been achieved by random guessing. In multiclass settings, 50% can already be very good. With our six genres, for instance, random guessing would result in only 16.7% (equal class sizes assumed).

The full training procedure now looks ...