Using of a matrix method of building of nonbinary decision trees for determining of stability of fixation of a tibia fracture

Abstract

Rapid development of information technologies, in particular, progress in methods of collection, storage and processing of data has allowed to collect huge data arrays with the purpose of their analysis in many organizations. Opportunities of experts are not enough because amount of these data are too much. This generates demand for methods of automatic data analysis number of which annually increase. Their application in biomechanical studies for the purpose of determining of stability of fixation of a fracture based on a physiological condition of patients is not exception. Most of algorithms for building of decision trees are time-consuming and poorly parallelized. Matrix method of building of nonbinary trees is presented in this article. Its advantages are a large reserve of an internal parallelism and lack of an inherent disadvantage of existing algorithms for building of decision trees - dependence on a choice of a variable of an initial split. Result of the method will be a complete set of rules on which can build a few trees with selection of the most significant of them in the future.