class of possible responses. This can be a simple 'yes or no' classification, or 'red, green or blue'. If you needed to predict whether an unknown person was male or female based on characteristics, you would choose this type of model. These are called discrete variables. Machine learning is a very technical space right now, and much of the cutting-edge work requires familiarity with linear algebra, calculus, math notation, and programming languages like Python. One of the things that helped me understand the overall flow

on an accessible level, however, was to think of machine learning models as applying weights to the characteristics of the data you feed them. The greater the functionality, the greater the weight. When reading about 'training models' it is helpful to visualize a chain ** fax number list** connected through the model to each weight, and when the model makes an estimate a cost function is used to tell you how much the guess was wrong and to gently, or harshly, pull the string in the direction of the correct answer, correcting all weights.

The part below gets a little technical with the terminology, so if that's too much for you, feel free to skip to the results and takeaways in the last section. Tackling Google Rankings Now that we had the data, we tried several approaches to solve the problem of predicting each web page's Google rank. Initially, we used a regression algorithm. That is, we set out to predict exactly how a site would rank for a given search term (e.g. a site will rank