Machine learning algorithms help perform tasks without extensive programming. These acquire this ability by learning from existing data.
At their core, MLA’s are intended to assist in making generalizations from observed data.
The more relevant the user data that’s provided to the machine learning algorithm, the better job it can do in the long run in finding the best products or offers to serve each specific user. For example, in addition to the user’s gender, the user’s past interactions with products (e.g., page-views and purchases) are often relevant for choosing which items to present to him/her.
Not all data about the user should be considered ‘relevant’ in a given situation. For example, if the algorithm is choosing between presenting the user with a promotion for a blue shirt or a promotion for a red shirt, data about the color of shirts the user has purchased in the past is clearly relevant, while data about the user’s shoe size could actually harm the performance of the algorithm.