A METHOD OF LIGHT WEIGHTED RANDOM FOREST CLASSIFICATION USING A SOFT TARGET LEARNING METHOD AND THE CLASSIFIER USING IT

摘要

The present invention relates to a method for classifying a lightweight random forest by applying a soft target learning method. The method for classifying a lightweight random forest comprises: a step (1) of learning a teacher random forest; a step (2) of extracting a probability value of each class; a step (3) of learning a student random forest; and a step (4) of performing classification by using the student random forest. The present invention can maintain the number of trees of a random forest while maintaining performance of the random forest, thereby being able to greatly reduce processing time and memory capacity.