Research on Intelligent Recognition Method of Risk Levels of Electricity Marketing Field Operation Types Based on Data Mining

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

Figure 7.

Figure 8.

Figure 9.

Figure 10.

Figure 11.

Figure 12.

Figure 13.

Abnormal behavior recognition performance

Influencing factor	Accuracy rate%	Recall rate%
Weak light	91.4	96.9
Weak light and sparse	95.7	96.4
Strong light	92.4	97
Strong light, sparse	97.3	96.4
Total	94.2	96.5

Test data for the power operation scenario

Test frame number	Correct number of action frames	Success rate
1086	1023	0.942

Training parameters

Training parameter	Parameter value
Training data set	35
Test data set	35
Learning rate	0.3
Iteration number	50

Abnormal behavior identification performance comparison

Method	Accuracy rate	Processing speed/FPS
Light flow method	90.52	8.82
Motion Instability	92.36	11.69
Support vector machine	80.63	16.8
Markov	83.55	15.06
This algorithm	93.87	18.33

The algorithm actually runs the data

	Total frame number	Inverted action frame number	Successful identification number	Success in identifying reverse times	Overall accuracy	Reverse recognition accuracy
1	342	98	334	95	0.928	0.965
2	262	75	248	78	0.936	0.966
3	428	167	387	155	0.932	0.963

The cause of the power operation accident

Accident category	Serial number	Precondition	Results	Confidence
Electrical shock (R1)	1	A13,D13	C22	0.902
	2	C22,D32	R1	0.991
	3	D22	D32	0.829
	4	F12,F13	A13	0.969
	5	D12	B21	0.993
	6	A13,D12	A33	0.92
	7	A33,D12	C13	0.944
	8	B21,C13	R1	0.997
	9	F14	C21	0.84
	10	A32,E13	R1	0.996
	11	A13,C21	A32	0.944
Object strike(R2)	12	A33,D41	C11	0.944
	13	A13	A33	0.815
	14	F12,F13	A13	0.964
	15	D14	B22	1.000
	16	A33,D14	C11	0.938
	17	B22,C11	R2	1.000
High fall(R3)	18	D21,D31	A22	0.959
	19	F14	C21	0.927
	20	C21,E14	B23	0.998
	21	A22,B23	R3	1.000
	22	F11	A12	0.901
	23	E22	B14	0.92
	24	A12	A51	0.983
	25	A51	B14	0.841
	26	D11,B14	R3	0.996
	27	D33,A13	A33	0.936
	28	F12,F13	A13	0.964
	29	A33,D11	C12	0.82
	30	C12,D11	R3	1.000

The accuracy and recall rate of three kinds of actions

	Precision	Recall
0.Sit down	0.71	0.81
1. Standing	0.94	0.86
2.Inversely	0.69	0.96

The cause of the power operation is a frequent collection

Accident category	Frequent set	Support
Electrical shock (R1)	{A13,C22,D13,D22,D32,F12,F13,R1}	0.185
	{A13,A33,B21,C13,D12,F12,F13,R1}	0.196
	{A13,A32,C21,E13,F12,F13,F14,R1}	0.221
Object strike(R2)	{A13,A33,B22,C11,D41,F12,F13,R2}	0.301
	{A13,A33,B22,C11,D14,F12,F13,R2}	0.337
High fall(R3)	{A22,B23,C21,D21,D31,E14,F14,R3	0.168
	{A12,A51,B14,D11,E22,F11,F12,R3}	0.184
	{A13,A33,C12,D11,D33,F12,F13,R3}	0.335

Data training results

Iteration rotation	Mean loss	Top1 accuracy	Top5 accuracy
10	4.6245	11.26%	32.76%
20	4.5156	14.45%	29.43%
30	3.8516	19.57%	42.72%
40	3.457	26.77%	48.67%
50	3.3162	25.66%	50.42%
60	3.562	31.22%	54.10%