Thus, the computer can learn from the data to make predictions from extracting the features from the data.

The training sample is often transformend into a string of numbers. And features are usually extracted from the numbers. Algorithms like nearest neighbor, linear regression, and logistic regression can be used to make predictions.

For example, in speech recognition, the computer can learn from the features in sound waves to recognize the speech.

waveform of audio

Waveform of "Duke"

mfcc extracted from audio

MFCC (a kind of feature extracted) of "Duke"