Machine learning is a broad term for a collection of mathematical techniques that learn patterns from data. In general, a machine learning algorithm will take a collection of data as an input and will output a model - an object that encodes the relationships describing the data. This model can then be used to predict the characteristics of new unseen data. For example, perhaps a telecommunications company would like to know the volume of traffic expected on its network tomorrow. A very simple model would be to predict that tomorrow's traffic would be the same as today's. This model is trained on only one datapoint and while it is not a very sophisticated model, it probably does a relatively good job. One could make a more sophisticated model by predicting traffic not based on just today's data, but perhaps taking an average of the data from the previous week. This may improve on the previous model by smoothing out daily inconsistencies. Going one step further, an even more powerful model might take in data from previous months or even years and take traffic growth trends and seasonal patterns into account when making predictions.
展开▼
