Electricity stealing analysis method and system based on rough set and deductive learning

The invention discloses an electricity larceny analysis method and system based on a rough set and deductive learning. The method comprises the steps of classification analysis, wherein whether label data in an electricity larceny analysis background is sufficient or not is divided into a supervised condition and an unsupervised condition; sample augmentation: learning the power utilization data set to obtain a group of new features, and augmenting the power utilization data training sample based on the features; data processing: carrying out classification processing on the original data, and comparing a result with a result obtained by domain knowledge to form a new corrected supervised sample; machine learning: obtaining an electricity larceny model hypothesis space through anti-deductive learning, performing machine learning in the hypothesis space to obtain a hypothesis model for judging electricity larceny, and checking the hypothesis model; data updating: performing deletion, modification and expansion