Table 6 Normalized feature importance and the selection of the top 10 most important features based on the three best-performing models.

Feature	Logistic Regression	SVM	Gradient Boosting	Decision Tree	Random Forest	KNN	Neural Network	Average [RF, LR, KNN]1
LAD
T2_13	0.00	0.73	0.47	0.66	1.00	1.00	1.00	0.66
SRS7	0.76	0.59	0.00	0.07	0.30	0.81	0.81	0.62
SRS1	0.68	0.46	0.51	0.39	0.37	0.59	0.59	0.55
T2_14	0.17	0.35	1.00	1.00	0.92	0.54	0.54	0.54
SCS13	0.52	0.01	0.05	0.02	0.20	0.56	0.56	0.43
SCS8	0.57	0.29	0.00	0.03	0.23	0.47	0.47	0.42
T2_global	1.00	1.00	0.42	0.35	0.26	0.00	0.00	0.42
SRS8	0.50	0.06	0.16	0.18	0.20	0.49	0.49	0.40
SCS1	0.52	0.01	0.08	0.08	0.33	0.31	0.31	0.39
T2_7	0.55	0.29	0.11	0.01	0.38	0.21	0.21	0.38
RCA
T2_9	0.00	1.00	0.61	0.54	0.89	1.00	1.00	0.96
T2_3	0.35	0.07	0.25	0.04	0.70	0.68	0.68	0.69
T1_9	0.51	0.08	0.20	0.16	0.66	0.45	0.45	0.52
T2_4	0.21	0.57	0.20	0.14	0.46	0.40	0.40	0.42
T2_15	0.27	0.41	0.02	0.14	0.36	0.42	0.42	0.40
T1_3	0.54	0.06	1.00	1.00	1.00	0.10	0.10	0.40
SCS9	0.56	0.04	0.18	0.19	0.83	0.18	0.18	0.40
T1_4	0.63	0.06	0.27	0.24	0.91	0.12	0.12	0.39
T1_10	0.39	0.11	0.07	0.08	0.31	0.33	0.33	0.32
T1_15	0.22	0.56	0.12	0.26	0.31	0.30	0.30	0.30
LCX
SCS11	1.00	0.95	0.88	0.69	0.78	0.64	0.64	0.81
T2_6	0.01	0.89	1.00	1.00	1.00	1.00	1.00	0.66
SLS5	0.76	0.28	0.60	0.19	0.61	0.44	0.44	0.60
IZS	0.07	0.89	0.74	0.07	0.92	0.54	0.54	0.51
T2_12	0.40	0.10	0.50	0.33	0.60	0.52	0.52	0.51
SCS16	0.58	0.10	0.60	0.14	0.59	0.29	0.29	0.49
SCS12	0.78	0.59	0.01	0.02	0.18	0.23	0.23	0.40
SRS11	0.41	0.53	0.01	0.01	0.37	0.35	0.35	0.38
SCS6	0.74	0.28	0.19	0.15	0.21	0.18	0.18	0.37
T1_5	0.14	0.80	0.21	0.10	0.56	0.41	0.41	0.37

IZS Infarct Zone Size, KNN K-Nearest Neighbors, SCS Segmental Circumferential Strain, SLS Segmental Longitudinal Strain, SRS Segmental Radial Strain, SVM Support Vector Machine.
The numbers presented in the tables are estimated using max-min standardization.
¹In the evaluation of the features impotence in the LAD domain, Random forest, Logistic regression and K nearest neighbor have the best performance and the average score estimate based on these models.
²In the evaluation of the features impotence in the LAD domain, Random forest, Neural Network and K nearest neighbor have the best performance and the average score estimate based on these models.
³In the evaluation of the features impotence in the LAD domain, Random forest, Neural Network and Logistic Regression have the best performance and the average score estimate based on these models.

Search