Skip to main content

Table 5 Comparison of accuracy of 5-fold cross validation based on unstructured data models

From: Clinical assistant decision-making model of tuberculosis based on electronic health records

 

1-Fold

2-Fold

3-Fold

4-Fold

5-Fold

Mean

STD

US-PTDM

0.9451 (0.94, 0.9502)

0.9485 (0.9436, 0.9534)

0.9434 (0.9382, 0.9486)

0.9431 (0.9379, 0.9483)

0.9452 (0.9401, 0.9503)

0.9451 (0.9433, 0.9469)

0.0021

Text-CNN

0.9284 (0.9226, 0.9342)

0.9205 (0.9144, 0.9266)

0.9111 (0.9047, 0.9175)

0.9151 (0.9089, 0.9213)

0.9175 (0.9113, 0.9237)

0.9185 (0.9128, 0.9242)

0.0065

GRU

0.9112 (0.9048, 0.9176)

0.9124 (0.9061, 0.9187)

0.9125 (0.9062, 0.9188)

0.9084 (0.9019, 0.9149)

0.9165 (0.9103, 0.9227)

0.9122 (0.9097, 0.9147)

0.0029

LSTM

0.9051 (0.8985, 0.9117)

0.9104 (0.904, 0.9168)

0.8916 (0.8846, 0.8986)

0.9088 (0.9024, 0.9152)

0.9074 (0.9009, 0.9139)

0.9047 (0.898, 0.9114)

0.0076

Bi_LSTM

0.9152 (0.909, 0.9214)

0.9135 (0.9072, 0.9198)

0.9158 (0.9096, 0.922)

0.9272 (0.9214, 0.933)

0.9185 (0.9124, 0.9246)

0.9180 (0.9133, 0.9227)

0.0054