Self-Learning Predictions are generated by interactions of self-learning nodes. There are two level of self-learning: self-learning by each and every node and self-learning through interactions of slef-learning nodes. The predictions by the network of self-learning enables the selection of the best variables and the selections of best prediction models for each and every records.

This network of self-learning nodes are based on Bayesian Statistics and Milorad's network of self-learning nodes. Therefore the models are probabilistic in nature and constructed using conditional probabilities. The network does not require any human intervention in mathematical processing. There is no limit in the type of variables it can use or the number of records. It also produce prediction at the fraction of the time used by supervised learning techniques.