Polawat Witoolkollachit. Semantic data mining between Thai medical chief complaint and ICD-10 from Thai national health data center (HDC-MOPH). Doctoral Degree(Information and Data Science (International Program)). King Mongkut's University of Technology North Bangkok. Central Library. : King Mongkut's University of Technology North Bangkok, 2021.
Semantic data mining between Thai medical chief complaint and ICD-10 from Thai national health data center (HDC-MOPH)
Abstract:
This study aims to extract medical-related concepts from the Chief Complaint (CC), in Thai language, to match with the information that appears in the database of the 10th revision of International Statistical Classification of Disease and Related Health Problems (ICD-10) in order facilitate medical processes such as double-checking, history gathering and diagnosis provisioning by machine learning. Supervised corpus is used for CC word segmentation and concepts tagging. From experiment result of 10-fold cross validation, Deep Learning (DL) algorithm apparently outperformed the other algorithms in all measurements. In terms of the precision score, DL obtained the best result at 0.72 while Naïve Baye (NB) and Decision Tree (DT) yielded similar precision score at 0.63-0.64. For the recall score, DL placed first at 0.73 and followed closely by DT with slightly lower score of 0.72. Overall, Deep Learning achieved the highest performance followed by Decision Tree, Naïve Baye, and Supporting Vector Machine. The algorithm Deep Learning performed better than the other chosen algorithms. The results thus present a clear fact that Deep Learning is most compatible with our CC algorithm design.