Preprocessing of the dataset using logarithm to the base 2 normalization to remove bias. It discretize continuous-valued attributes by taking logarithm to the base 2 and then converting to integer. This is done for each attribute value z using the computation: if (z > 2)z = _(log2(z)) + 1. Before taking the logarithm, the attributes that take fractional values in the range [0, 1] are multiplied by 100 so that they take values in the range [0, 100].
Use of a heuristic method to identify appropriate range for δj and δaj by exhaustive experiments using benchmark and real-life datasets.