Improving Numerical Reasoning Capabilities of Inductive Logic Programming Systems

Abstract

Inductive Logic Programming (ILP) systems have been largely applied to classification problems with a considerable success. The use of ILP systems in problems requiring numerical reasoning capabilities has been far less successful. Current systems have very limited numerical reasoning capabilities, which limits the range of domains where the ILP paradigm may be applied.This paper proposes improvements in numerical reasoning capabilities of ILP systems. It proposes the use of statistical-based techniques like Model Validation and Model Selection to improve noise handling and it introduces a new search stopping criterium based on the PAC method to evaluate learning performance.We have found these extensions essential to improve on results over statistical-based algorithms for time series forecasting used in the empirical evaluation study.KeywordsRoot Mean Square ErrorBayesian Information CriteriumMinimum Description LengthModel Selection CriterionInductive Logic ProgrammingThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.