An open-source tool to identify active travel from hip-worn accelerometer, GPS and GIS data

Table 3 The confusion-matrix and accuracy scores per mode, expressed as percentages, compared with manually-identified data

		Observed mode					Mode	Positive predictive value^a	Sensitivity^b	F1 score^c
		Cycle	Walk	Train	Vehicle	Stationary	Mode	Positive predictive value^a	Sensitivity^b	F1 score^c
Predicted mode	Cycle	3651	1426	116	215	385	Cycle	63.0	94.3	75.5
	Walk	15	23,346	12	128	1858	Walk	92.1	75.6	83.0
	Train	43	749	6275	89	1318	Train	74.1	96.6	83.8
	Vehicle	156	585	38	23,684	3726	Vehicle	84.0	97.1	90.1
	Stationary	8	4776	58	285	329,807	Stationary	98.5	97.8	98.2

^aPositive predictive value (PPV) or Precision, the ratio of true positives to the sum of true and false positives
^bSensitivity or Recall/ True positive rate/Detection rate, the ratio of true positive to true positives and false negatives
^cF1 score, the harmonic mean of PPV and sensitivity

ISSN: 1479-5868