We present an intelligent system for VKF-method based on a Markov chain approach to generation of hypotheses about causes of presence/absence of effect under study. The system uses coupling Markov chains that terminate with probability 1. Since each hypothesis is generated by an independent run of the Markov chain, the system makes the induction step in parallel by several threads. After that the abduction step refines the hypotheses by the CloseByOne operation with training examples (in several threads too). Then the system predicts presence/ absence of the effect by the analogical reasoning.We test the system on SPECT dataset from UCI machine learning repository. The accuracy is 85.56 percent (that exceeds 84.0 percent accuracy of the CLIP3 algorithm developed by the authors of the dataset).
CITATION STYLE
Vinogradov, D. V. (2014). VKF-method of hypotheses generation. Communications in Computer and Information Science, 436, 237–248. https://doi.org/10.1007/978-3-319-12580-0_25
Mendeley helps you to discover research relevant for your work.