ENHANCE DECISION TREE ALGORITHM FOR UNBALANCED DATA: RAREDTREE

Unbalanced data classification is important when misclassification rate of rare instances is huge. Medical diagnosis field is an example. Existing techniques of unbalanced data classification are based on sampling techniques which suffer from overlapping and increase learning time. To develop effective intelligent system for domain of unbalanced data, minority example should be classified with good accuracy. But traditional machine learning algorithms lack this features and they are biased towards majority class. We proposed new optimal algorithm based on decision tree algorithm. We modified the decision tree algorithm and developed new algorithm called RareDTree which classify minority instances with good accuracy without compromising the accuracy of majority class instances. RareDTree also eliminate the need of data sampling.