|
چکیده
|
One of the challenges of artificial intelligence and data mining algorithms in the automatic diagnosis of diseases is imbalanced dataset problem. The lack of data balancing will reduce accuracy of the results, which is very dangerous in diseases like breast cancer. This paper presents an algorithm for balancing number of instances in breast cancer datasets. The proposed algorithm uses ReliefF for weighting and ranking of instances. ReliefF is a well-known algorithm for ranking features, but, here, we used it with some modifications to rank the instances. After ranking the instances, based on the weight obtained, a combination of undersampling and oversampling methods is used to balance the dataset. The obtained results from testing the proposed algorithm on two datasets show the effectiveness of this algorithm.
|