Word2Vec inversion and traditional text classifiers for phenotyping lupus

Table 5 Top 25 CUIs according to the variable importance extracted from scikit-learn’s ExtraTreesClassifier [22]

Rank	CUI	Description	VIMP
1	C0042014	Laboratory: Urine Examination	0.0307
2	C0699177	Plaquenil	0.0258
3	C0024141	Systemic Lupus Erythematosus	0.0236
4	C0194073	Kidney Biopsy	0.0208
5	C0024204	Lymph Node	0.0179
6	C0008031	Nonspecific Chest Pain	0.0166
7	C0018966	Heme	0.0158
8	C2711450	Enlargement (Morphological Anomaly)	0.01502
9	C0014597	Epithelial Cell	0.0111
10	C0023516	Leukocytes	0.0100
11	C0003243	Antinuclear Antibody (ANA)	0.0094
12	C0002170	Alopecia	0.0089
13	C0024202	Lymph	0.0085
14	C1267547	Entire Mouth Region	0.0084
15	C0009780	Connective Tissue	0.0083
16	C0229671	Serum	0.0068
17	C0042036	Urine	0.0065
18	C0014060	St. Louis Encephalitis	0.0062
19	C0038999	Swelling	0.0061
20	C1269549	Entire Zygoma	0.0060
21	C0036749	Serositis	0.0060
22	C0033684	Proteins	0.0059
23	C0014239	Endoplasmic Reticulum	0.0059
24	C0009782	Connective Tissue Disorder	0.0058
25	C0024143	Lupus Nephritis	0.0058

CUI descriptions were extracted from MetamorphoSys [7]. A graph of the degradation of variable importance for these CUIs can be found in Fig. 4

ISSN: 1472-6947