Research Article

Natural Language Processing Based Instrument for Classification of Free Text Medical Records

Table 2

Results of SVM versus KNN (feature selection classic method).

Feature selection classic method
Class nameSubclass nameSVMKNN
AccERRAccERR

Ultrasonography0,910,800,850,850,150,830,810,820,840,17
Liver0,750,530,620,860,140,720,510,600,850,15
Biliary system0,570,440,500,780,220,560,400,460,760,24
Kidney and urinary system0,940,810,870,970,030,780,590,670,920,08
Gynecologic0,910,870,890,960,040,700,630,660,870,13
Thyroid0,830,790,810,940,060,810,710,760,910,09
Breast0,900,770,830,960,040,870,590,700,920,08
Vascular Doppler0,870,720,790,950,050,870,700,770,950,05

X-ray0,890,810,850,850,150,810,790,800,810,19
Chest0,890,850,870,960,040,780,840,810,940,06
Abdomen0,860,760,810,920,080,760,640,700,870,13
Spine0,890,770,830,930,070,780,630,700,880,12
Limbs0,990,770,870,970,040,830,570,680,910,09
Esophagus and stomach0,910,880,890,950,050,690,640,660,840,16
Large and small bowels0,910,940,930,980,020,840,660,740,940,07

Endoscopy 0,970,800,880,980,020,930,750,830,970,03