Inductive logic programming approach analysis in bipolar disorder - pilot study
Nenhuma Miniatura disponível
Citações na Scopus
Tipo de produção
conferenceObject
Data de publicação
2013
Título da Revista
ISSN da Revista
Título do Volume
Editora
WILEY-BLACKWELL
Autores
SALVINI, R.
MADUREIRA, D. Quintela Mendes
SCIPPA, A. M.
KAPCZINSKI, F.
Citação
BIPOLAR DISORDERS, v.15, suppl.1, Special Issue, p.69-69, 2013
Resumo
Objective: Inductive Logic Programming (ILP) is a modern approach for machine learning and pattern recognition based in logic systems for knowledge representation. ILP solve problems involving multi-relational learning, processing of data carried out by different conventional statistical techniques, not limited by the statistical restrictions. The so criticized phenomenon of ‘black box’ does not occur, since the ILP system answers a question of interest by describing relations (rules) extracted from the data set mined. Interpretations of the rules exemplified occur naturally, intuitively and induce processes. Due to the multi-relational aspect of Bipolar Disorder, it is reasonable to think that ILP could be seen as a method applied in the research field. This is a pilot study to analyze the precision performance of ILP in BD research. Methods: The demographic and clinical data (108 variables) from 600 patients (183 men and 417 women) from the Brazilian Bipolar Research Network were compared considering gender and presence of rapid cycling in the last year initially through biostatistics techniques and with assessments of ILP (Aleph method) using two approaches: considering all variables and just statistic significant variables founded through biostatistics. The ILP performance (all rules together) was measured evaluating the contingence table considering true and false positive and negative cases. Results: The gender evaluation considering all variables generated 93 rules (41 for men and 52 for women, 98.2% and 92.4% coverage respectively), and 89 rules considering with just significant statistic variables (33 for men and 56 for women, 88.8% and 65.9% cover-age respectively). The observed performance range to identify women and men were: accuracy 89.7% to 98.8%, balanced accuracy 82.9% to 99.1%; sensibility 65.9% to 98,2% and F-measure 79.4% to 99.1%. No range was observed for specificity and precision, both 100%. The rapid cycling evaluation considering all variables generated 69 rules (26 rapid cyclers and 43 no rapid cyclers, 89.7% and 99.6% coverage respectively) and 76 rules (28 rapid cyclers and 48 no rapid cyclers, 92.8% and 98.7% coverage respectively) considering with just significant statistic variables. The observed performance range to identify rapid cyclers were: accuracy 98.2% to 98.9%, balanced accuracy 94.8% to 99.8%; sensibility 89.7%% to 99.6% and F-measure 94.6% to 99.8%. No range was observed for specificity and precision, both 100%. In both analyses were observed rules with unrelated significant statistically variables presenting clinical relevance. Discussion: The ILP approach results showed a high precision to identify BD patients considering gender and presence of rapid cycling in the last year and described several associations opening space for a more systematic analysis. This is mainly due to the relational intrinsic nature if ILP. More studies with larger and more complex data are necessary to prove the validity to use ILP in BD research.
Palavras-chave
bipolar disorder, methodology, inductive logic programming, exploratory