Information Extraction and Prediction Using Partial Keyword Combination and Blends Measure
L. Sathish Kumar, A. Padmapriya · IETE Journal of Research · 2018
Electronic medical records (EMRs) contain clinical information about individuals. Nowadays, EMRs are widely used because of reusability and to reduce the costs associated with sharing the clinical information in healthcare sectors. Most of the EMRs are in plain text and unstructured form, hence it is difficult to filter and retrieve useful knowledge from the records. The knowledge acquired from EMR is effectively utilized by the Healthcare stakeholders. The proposed method extracts information, especially the information about diseases, using ICD-10 code which is supplemented with the statistical evidence. Here extraction of information is in the form of query combination of keywords. With the help of this method, we could generate predictions for the diseases listed in ICD-10. The predictions are based on standard data-sets from Universal Classification Irvin (UCI) repository, Word Health Organization, and EMR data-sets. The proposed method of Information Extraction and Prediction using Partial Keyword Combination and Blends Measure (IEPKCB) yields 97.4% precision, 95.0% recall, and 96.1% F-measure. The average execution time of IEPKCB is fewer than 0.84 seconds.