BeeMachine model performance

Species list and performance metrics

Species	Precision	Recall	Error rate	F1-score	Num test images	Num training images
Bombus affinis	0.96	0.96	0.04	0.96	263	1035
B. appositus	0.88	0.90	0.10	0.89	162	715
B. auricomus	0.87	0.92	0.08	0.90	353	1369
B. bifarius	0.94	0.92	0.08	0.93	986	4037
B. bimaculatus	0.89	0.91	0.09	0.90	1089	4352
B. borealis	0.93	0.93	0.07	0.93	416	1621
B. caliginosus	0.55	0.68	0.32	0.61	34	109
B. centralis	0.80	0.88	0.12	0.84	318	1409
B. citrinus	0.86	0.92	0.08	0.89	227	911
B. crotchii	0.87	0.83	0.17	0.85	63	266
B. cryptarum	0.94	0.92	0.08	0.93	74	316
B. flavidus	0.78	0.83	0.17	0.81	197	742
B. fervidus	0.91	0.86	0.14	0.88	849	3430
B. flavifrons	0.86	0.84	0.16	0.85	730	3007
B. fraternus	0.83	0.96	0.04	0.89	70	315
B. frigidus	0.92	0.82	0.18	0.87	28	107
B. griseocollis	0.94	0.96	0.04	0.95	1389	5277
B. huntii	0.93	0.91	0.09	0.92	652	2427
B. impatiens	0.97	0.93	0.07	0.95	2020	7988
B. insularis	0.77	0.93	0.07	0.84	204	977
B. melanopygus	0.94	0.93	0.07	0.93	913	3575
B. mixtus	0.88	0.91	0.09	0.90	586	2409
B. morrisoni	0.89	0.92	0.08	0.91	52	211
B. nevadensis	0.95	0.93	0.07	0.94	380	1478
B. occidentalis	0.96	0.91	0.09	0.93	188	726
B. pensylvanicus	0.95	0.95	0.05	0.95	1172	4539
B. perplexus	0.91	0.93	0.07	0.92	468	1827
B. rufocinctus	0.85	0.77	0.23	0.81	789	3098
B. sandersoni	0.55	0.63	0.37	0.59	38	111
B. sitkensis	0.68	0.70	0.30	0.69	109	469
B. sylvicola	0.62	0.74	0.26	0.67	61	254
B. ternarius	0.94	0.95	0.05	0.94	1064	4453
B. terricola	0.94	0.94	0.06	0.94	360	1481
B. vagans	0.79	0.85	0.15	0.82	520	2152
B. vandykei	0.75	0.72	0.28	0.73	81	340
B. vosnesenskii	0.96	0.96	0.04	0.96	1051	4287

Confusion matrix

A confusion matrix shows the relationship between true and predicted values. For example, with 2020 test images, B. impatiens was correctly classified 1888 times. It was most frequently confused with B. bimaculatus (54 times) and B. griseocollis (43 times). Likewise, the species most often confused for B. bimaculatus and B. griseocollis was B. impatiens (21 and 13 times respectively).