Proposal of a simple recommendation system for small and medium enterprises for decision making based on unsupervised learning

Abstract

Recommendation systems are generally complicated, due they search to increase their reach and robustness, they combine different artificial intelligence approaches mainly of supervised learning. A disadvantage of this type of systems is that they must have a prior classification to be able to train a system and after they can be able to make decisions in a simmilar way that a human would do it; however, the task of classification is often expensive because is needed to consult with experts the possible classification (also known as label) that should be given to a specific data; although this method can be profitable for large companies, it is not for small and medium companies. This is the reason which the present work shows a proposal of a simple system that does not need to have a previous classification, allowing it to be profitable for small and medium enterprises in decision making.