Machine learning systems and methods

摘要: A method for using machine learning to solve problems having either a “positive” result (the event occurred) or “negative” did not occur), in which the probability of positive is very low and consequences are significant. Training data obtained subset that distilled application system. The training includes some records corresponding result, nearest neighbors from negative other result. system uses co-evolution approach obtain rule set predicting results after number cycles. fitness function derived use with type problem, such as based on sensitivity predictive value rules. rules validated entire data.