Fig. 3
From: An interpretable risk prediction model for healthcare with pattern attention
Case study. The figure shows the observed variables of a case patient during the last 24 h observation window (the hold-off window is 10 h). The black stars represent observed abnormal values with instance-wise contribution risks generated by RETAIN, while the colored squares are medical event patterns detected by PAVE. The events sharing the same colors are detected patterns. The sizes of black stars and colored squares denote the corresponding values of contribution risks. Note that only the events or patterns with relatively higher contribution risks are marked in the figure. PAVE found three high-risk patterns for the patient: (1) high SysBP and high temperature in orange squares; (2) high heart rate and high temperature in red squares; (3) stable high heart rate and high respiratory rate in blue squares