Decoration
max planck institut
informatik
mpii logo Minerva of the Max Planck Society
 

AMIE: Association Rule Mining under Incomplete Evidence in Ontological Knowledge Bases

PCA vs Standard Confidence

Based on our experimental setup, we compared the suitability of the standard and PCA confidence for prediction of facts. The following chart compares the aggregated number of new facts or predictions against the aggregated precision (the number of hits or correct predictions) when the rules are ranked by standard confidence and PCA confidence. It is important to remark that ranking by PCA confidence produces much more predictions because it is less conservative than the standard confidence, identifying potentially incomplete relations. The approximately 400K predictions were produced by the top 30 most PCA-confident rules, whereas the top 46 rules ranked by standard confidence produced less than 200K predictions. The PCA confidence also achieves better precision.

Standard vs PCA confidence comparison

[Back to results]

Our current efforts are focused on improving the precision of AMIE by taking into account schema information like data types, domains and ranges.