Combining structured and unstructured data for predictive models: a deep learning approach

Table 5 P value matrix of various model performances (AUROC) for long length of stay prediction. U, T, S represents unstructured data, temporal signals, and static information respectively

		LR			RF			Fusion-CNN			Fusion-LSTM
		T + S	U	U + T + S	T + S	U	U + T + S	T + S	U	U + T + S	T + S	U	U + T + S
LR	\(T + S\)	1	0.5643	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
	U	0.5643	1	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	0.0024	< 0.001	< 0.001	< 0.001	< 0.001
	\(U + T + S\)	< 0.001	< 0.001	1	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	0.0073
RF	\(T + S\)	< 0.001	< 0.001	< 0.001	1	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
	U	< 0.001	< 0.001	< 0.001	< 0.001	1	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
	\(U + T + S\)	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	1	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
Fusion-CNN	\(T + S\)	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	1	< 0.001	< 0.001	< 0.001	0.1134	< 0.001
	U	< 0.001	0.0024	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	1	< 0.001	< 0.001	< 0.001	< 0.001
	\(U + T + S\)	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	1	< 0.001	< 0.001	0.002
Fusion-LSTM	\(T + S\)	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	1	< 0.001	< 0.001
	U	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	0.1134	< 0.001	< 0.001	< 0.001	1	< 0.001
	\(U + T + S\)	< 0.001	< 0.001	0.0073	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	0.002	< 0.001	< 0.001	1

ISSN: 1472-6947