Member-only story
Whenever embarking upon machine learning and endeavouring to attain the best rate of accuracy possible, it is important to utilise a variety of models and methodologies in which those models are used.
One such mechanism that is used to attain better accuracy in making predictions is to normalise the data. When data is normalised, it is rescaled to a specific value, normally from a range of zero to one, with zero being the lowest value and one being the highest value in the dataset.
Machine learning algorithms tend to perform better or converge faster if the data they are training on is of a smaller scale. It is for this reason that programmers tend to normalise the data before training machine learning models on an estimator.
Normalisation also makes the process less sensitive to the scale of the features, thus attaining better coefficients after training:-
It is easy to become complacent and just normalise data as a matter of course, but it is important to train data that has not been normalised and then train it on the same data that has been normalised, selecting the model that achieves the best accuracy.
