Vision-Based Deep Q-Learning Network Models to Predict Particulate Matter Concentration Levels Using Temporal Digital Image Data

Table 6

Simulation study: sensitivity (Sen), specificity (Spe), Youden index, and standard errors in parentheses with an application to various video sequences and environmental conditions. The simulation was repeated 50 times.

# of frames	5 frames	10 frames	15 frames	20 frames
	Windiness (Use of fan)
Sen	0.8411 (0.0408)	0.8495 (0.0452)	0.9152 (0.0335)	0.9116 (0.0304)
Spe	0.8796 (0.0299)	0.9141 (0.0266)	0.9411 (0.0221)	0.9419 (0.0200)
Youden	0.7207 (0.0377)	0.7636 (0.0526)	0.8563 (0.0738)	0.8535 (0.0496)
	High Temperature (40°C)
Sen	0.8664 (0.0291)	0.8866 (0.0315)	0.8844 (0.0357)	0.9090 (0.0291)
Spe	0.9160 (0.0236)	0.9524 (0.0201)	0.9601 (0.0157)	0.9587 (0.0170)
Youden	0.7824 (0.0336)	0.8390 (0.0398)	0.8445 (0.0427)	0.8677 (0.0371)
	High Humidity (50%)
Sen	0.8758 (0.0274)	0.9317 (0.0221)	0.9478 (0.0193)	0.9392 (0.0234)
Spe	0.9562 (0.0185)	0.9647 (0.0199)	0.9961 (0.0029)	0.9821 (0.0146)
Youden	0.8320 (0.0312)	0.8764 (0.0298)	0.9439 (0.0209)	0.9213 (0.0308)
	High luminous Intensity (250lx)
Sen	0.8990 (0.0276)	0.9290 (0.0247)	0.9259 (0.0251)	0.9211 (0.0278)
Spe	0.8669 (0.0308)	0.8947 (0.0307)	0.9071 (0.0293)	0.9265 (0.0278)
Youden	0.7659 (0.0342)	0.8237 (0.0396)	0.8330 (0.0404)	0.8476 (0.0430)