Denormalize
Package
weka.filters.unsupervised.instance
Synopsis
Coming in Weka 3.7.2
An instance filter that collapses instances with a common grouping ID value into a single instance. Useful for converting transactional data into a format that Weka's association rule learners can handle. IMPORTANT: assumes that the incoming batch of instances has been sorted on the grouping attribute. The values of nominal attributes are converted to indicator attributes. These can be either binary (with f and t values) or unary with missing values used to indicate absence. The later is Weka's old market basket format, which is useful for Apriori. Numeric attributes can be aggregated within groups by computing the average, sum, minimum or maximum.
Available in Weka 3.7.1. Available via the package management system for Weka >= 3.7.2 (denormalize).
Options
The table below describes the options available for Denormalize.
Option |
Description |
---|---|
aggregationType |
The type of aggregation to apply to numeric attributes. |
groupingAttribute |
Set the attribute that defines the groups (e.g. transaction ID). |
useOldMarketBasketFormat |
Output instances that contain unary attributes with absence indicated by missing values. Apriori operates faster with this format. |
useSparseFormat |
Output sparse instances (can't be used in conjunction with useOldMarketBasketFormat). |
Capabilities
The table below describes the capabilities of Denormalize.
Capability |
Supported |
---|---|
Class |
String class, Binary class, No class, Date class, Nominal class, Numeric class, Unary class, Relational class, Empty nominal class |
Attributes |
String attributes, Date attributes, Unary attributes, Numeric attributes, Binary attributes, Nominal attributes, Missing values, Empty nominal attributes |
Min # of instances |
0 |