Filtering free-text medical data based on machine learning

Abstract

This article describes the results of data filtering of electronic health records for patients diagnosed with aortic aneurysm in two different medical centers to prepare data for further feature extraction. The accuracy improvement of filtered data was achieved by using machine learning methods of classification and natural language processing methods, taking into account the specificity of Russian language. Based on accuracy and F-measure, two methods of data filtering were compared: 1) rule-based approach; 2) classification approach. The results show that the designed classification is appropriate in terms of accuracy for data filtering.

Ähnliche Arbeiten