Supervised Classification of Healthcare Text Data Based on Context-Defined Categories

Achieving a good success rate in supervised classification analysis of text dataset, where the relationship between and its label can be extracted from context, but not isolated words text, is still an important challenge facing fields statistics machine learning. For this purpose, we present novel mathematical framework. We then conduct comparative study established methods for case corresponding clearly depicted by specific text. In particular, use logistic LASSO, artificial neural networks, support vector machines, decision-tree-like procedures. This methodology applied to real involving mapping Consolidated Framework Implementation Research (CFIR) constructs health-related data achieves prediction over 80% when just first 55% or more, used training remaining testing. The results indicate that useful accelerate CFIR coding process.