Discretization techniques can be used to convert continuous attributes to nominal attributes. In this way, the number of values for a given continuous attribute is reduced by dividing the attribute into a range of values. Actual data values are replaced with interval value labels.
Machine-learning algorithms are typically recursive; to process large amounts of data a great deal of time is spent to sort the data at every step. It is clear that the smaller the number of distinct values to be ordered, the faster these methods should be. That is why these techniques are particularly beneficial.
In discretization, the raw values of a numeric attribute are replaced by labels or conceptual labels. For example, the continuous value of the measured temperature in one day can be divided into three bins (0-10, 11-20, 21-30) or can be divided into the three conceptual...
