Hitachi Vantara Pentaho Community Wiki
Child pages
  • MITI
Skip to end of metadata
Go to start of metadata

Package

weka.classifiers.mi

Synopsis

MITI (Multi Instance Tree Inducer): multi-instance classification based a decision tree learned using Blockeel et al.'s algorithm. For more information, see

Hendrik Blockeel, David Page, Ashwin Srinivasan: Multi-instance Tree Learning. In: Proceedings of the International Conference on Machine Learning, 57-64, 2005.

Luke Bjerring, Eibe Frank: Beyond Trees: Adopting MITI to Learn Rules and Ensemble Classifiers for Multi-instance Data. In: Proceedings of the Australasian Joint Conference on Artificial Intelligence, 2011.

From multiInstanceLearning package version 1.0.3 for Weka >= 3.7.2.

Options

The table below describes the options available for MITI.

Option

Description

attributesToSplit

The number of randomly chosen attributes to consider for splitting.

b

Whether to use bag-based statistics for estimates of proportion.

ba

Multiplier for count influence of a bag based on the number of its instances.

debug

If set to true, classifier may output additional info to the console.

k

The value used in the tozero() method.

l

Whether to scale based on the number of instances.

seed

The random number seed to be used.

splitMethod

The method used to determine best split: 1. Gini; 2. MaxBEPP; 3. SSBEPP

topNAttributesToSplit

Value of N to use for top-N attributes to choose randomly from.

unbiasedEstimate

Whether to used unbiased estimate (EPP instead of BEPP).

Capabilities

The table below describes the capabilities of MITI.

Capability

Supported

Class

Missing class values, Binary class

Attributes

Unary attributes, Numeric attributes, Binary attributes, Relational attributes, Nominal attributes, Date attributes, String attributes, Empty nominal attributes

Other

Only multi-Instance data

Min # of instances

1

  • No labels