Data preprocessing in WEKA

Data preprocessing includes processing of missing data values, standardization, standardization, and discretization.

Processing of missing data values: WEKA. Filters. unsupervised. Attribute. replacemissingvalues.For the value attribute, use the average value instead of the missing value. For the nominal attribute, use its mode (the most frequently seen value) instead of the missing value.

Standardization: Class WEKA. Filters. unsupervised. Attribute. Standardize. Normalize the values of all numerical attributes in a given dataset to a normal distribution of 0 mean and unit variance.

Standardization: Class WEKA. Filters. unsupervised. Attribute. normalize. Normalize All numeric attribute values in a given dataset, except for class attributes. The result value is in the range [0, 1] by default. However, using the zoom and shift parameters, we can normalize the value attribute value to any range. For example, when scale = 2.0 and translation =-1.0, You can normalize the attribute value to the range [-1, + 1].

Discretize: Class WEKA. Filters. Supervised. Attribute. discretize and WEKA. Filters. unsupervised. Attribute. discretize. Discretization of supervised and unsupervised numeric attributes is used to separate numeric attributes in a dataset to Classification attributes.

Reprinted from: http://blog.csdn.net/hunauchenym/article/details/5847314