Challenge@RuleML2015 EasyMiner/R Preview: Towards a Web Interface for Association Rule Learning and...
Transcript of Challenge@RuleML2015 EasyMiner/R Preview: Towards a Web Interface for Association Rule Learning and...
EasyMiner/R Preview Towards a Web Interface for Association Rule Learning and Classification in R
Stanislav Vojíř, Václav Zeman, Jaroslav Kuchař*, Tomáš Kliegr
Faculty of Informatics and Statistics
University of Economics, Prague
Czech Republic
* Also affiliated with
Web Intelligence Research Group
Faculty of Information Technology
Czech Technical University in Prague
What is association rule learning?
Association rules
learner
Parameters:
Minimum confidence: 90%
Minimum support: 20%
because:
one transaction out of five contains butter, bread and milk
support is 1/5=20%
all transactions which contain butter and bread contain milk
confidence is 1/1=100%
Example adapted from https://en.wikipedia.org/wiki/Association_rule_learning
EasyMiner predecessor
(RuleML 2010 Challenge)
Current version (2015)
Features of EasyMiner/R
Discovers association rules in given dataset
Interactive discovery
Found rules are editable
Save discovered rules to (business rules) knowledge base
Create classification models
Fully automatic
Human editable rule set
Less rules with built-in rule pruning
Evaluation
EasyMiner offers two backends: LISp-Miner and the arules
package from R
LISp-Miner offers many advanced features, including dynamic
binning during mining and refined ways of constraining the search
space
When these features are not required, the R arules backend is
faster, especially on larger datasets
Evaluation on a dataset generated for the ESWC 2014 Recommender Systems
Challenge (72,371 rows, 7 attributes)
Live demo
http://easyminer.eu