Tree-based models for inductive classification on the Web Of Data

摘要: Abstract The Web of Data, which is one the dimensions Semantic (SW), represents a tremendous source information, motivates increasing attention to formalization and application machine learning methods for solving tasks such as concept learning, link prediction, inductive instance retrieval in this context. However, Data also characterized by various forms uncertainty, owing its inherent incompleteness (missing uneven data distributions) noise, may affect open distributed architectures. In paper, we focus on task regarded classification problem. proposed solution framework Terminological Decision Trees from examples described an ontological knowledge base, be used performing classifications. For purpose, suitable pruning strategies new prediction procedure are proposed. Furthermore, order tackle class-imbalance distribution problem, extended ensembles called Random Forests . has been evaluated, comparative experiments, with main state art solutions grounded similar approach, showing that: (1) employment formalized can improve model predictiveness; (2) outperform usage single Tree , particularly when base endowed large number concepts roles; (3) exploited related problems, predicting values given properties finite ranges.