Research on Spectral Recognition of Drug Mixture Based on SVM-MLP Fusion Model

Wenjie Yan; Wenhui Lu; Jifen Wang

doi:10.3788/LOP202158.1404003

Journals >Laser & Optoelectronics Progress >Volume 58 >Issue 14 >Page 1404003 > Article

Laser & Optoelectronics Progress
Vol. 58, Issue 14, 1404003 (2021)

Research on Spectral Recognition of Drug Mixture Based on SVM-MLP Fusion Model

Wenjie Yan¹, Wenhui Lu², and Jifen Wang^1、*

Author Affiliations

¹School of Investigation, People's Public Security University of China, Beijing 102600, China

²Henan Police College, Zhengzhou, Henan 450000, China

show less

DOI: 10.3788/LOP202158.1404003 Cite this Article Set citation alerts

Wenjie Yan, Wenhui Lu, Jifen Wang. Research on Spectral Recognition of Drug Mixture Based on SVM-MLP Fusion Model[J]. Laser & Optoelectronics Progress, 2021, 58(14): 1404003 Copy Citation Text

show less

Fig. 1. Distribution of contribution rate of each component

Download full size

Fig. 2. Schematic diagram of support vector machine

Download full size

Fig. 3. Accuracy comparison chart

Download full size

Fig. 4. MLP simple structure schematic diagram

Download full size

Fig. 5. Classification effects diagram of two models under different training ratios

Download full size

Samples	Heroin quality /mg	Additives	Additive quality /mg	Heroin mass score
1	0.5	Caffeine	4.5	0.1
2	0.5	Glucose	4.5	0.1
3	0.5	Phenacetin	4.5	0.1
4	0.5	Starch	4.5	0.1
5	0.5	Sucrose	4.5	0.1
6	1.0	Caffeine	4.0	0.2
7	1.0	Glucose	4.0	0.2
8	1.0	Phenacetin	4.0	0.2
9	1.0	Starch	4.0	0.2
10	1.0	Sucrose	4.0	0.2
11	1.5	Caffeine	3.5	0.3
12	1.5	Glucose	3.5	0.3
13	1.5	Phenacetin	3.5	0.3
14	1.5	Starch	3.5	0.3
15	1.5	Sucrose	3.5	0.3
16	2.0	Caffeine	3.0	0.4
17	2.0	Glucose	3.0	0.4
Samples	Heroin quality /mg	Additives	Additive quality /mg	Heroin mass score
18	2.0	Phenacetin	3.0	0.4
19	2.0	Starch	3.0	0.4
20	2.0	Sucrose	3.0	0.4
21	2.5	Caffeine	2.5	0.5
22	2.5	Glucose	2.5	0.5
23	2.5	Phenacetin	2.5	0.5
24	2.5	Starch	2.5	0.5
25	2.5	Sucrose	2.5	0.5
26	3.0	Caffeine	2.0	0.6
27	3.0	Glucose	2.0	0.6
28	3.0	Phenacetin	2.0	0.6
29	3.0	Starch	2.0	0.6
30	3.0	Sucrose	2.0	0.6
31	3.5	Caffeine	1.5	0.7
32	3.5	Glucose	1.5	0.7
33	3.5	Phenacetin	1.5	0.7
34	3.5	Starch	1.5	0.7
35	3.5	Sucrose	1.5	0.7
36	4.0	Caffeine	1.0	0.8
37	4.0	Glucose	1.0	0.8
38	4.0	Phenacetin	1.0	0.8
39	4.0	Starch	1.0	0.8
40	4.0	Sucrose	1.0	0.8
41	4.5	Caffeine	0.5	0.9
42	4.5	Glucose	0.5	0.9
43	4.5	Phenacetin	0.5	0.9
44	4.5	Starch	0.5	0.9
45	4.5	Sucrose	0.5	0.9
46	0.5	Caffeine	4.5	0.1
47	0.5	Glucose	4.5	0.1
48	0.5	Paracetamol	4.5	0.1
49	0.5	Phenacetin	4.5	0.1
50	0.5	Starch	4.5	0.1
51	1.0	Caffeine	4.0	0.2
52	1.0	Glucose	4.0	0.2
53	1.0	Paracetamol	4.0	0.2
54	1.0	Phenacetin	4.0	0.2
55	1.0	Starch	4.0	0.2
56	1.5	Caffeine	3.5	0.3
57	1.5	Glucose	3.5	0.3
58	1.5	Paracetamol	3.5	0.3
Samples	Heroin quality /mg	Additives	Additive quality /mg	Heroin mass score
59	1.5	Phenacetin	3.5	0.3
60	1.5	Starch	3.5	0.3
61	2.0	Caffeine	3.0	0.4
62	2.0	Glucose	3.0	0.4
63	2.0	Paracetamol	3.0	0.4
64	2.0	Phenacetin	3.0	0.4
65	2.0	Starch	3.0	0.4
66	2.5	Caffeine	2.5	0.5
67	2.5	Glucose	2.5	0.5
68	2.5	Paracetamol	2.5	0.5
69	2.5	Phenacetin	2.5	0.5
70	2.5	Starch	2.5	0.5
71	3.0	Caffeine	2.0	0.6
72	3.0	Glucose	2.0	0.6
73	3.0	Paracetamol	2.0	0.6
74	3.0	Phenacetin	2.0	0.6
75	3.0	Starch	2.0	0.6
76	3.5	Caffeine	1.5	0.7
77	3.5	Glucose	1.5	0.7
78	3.5	Paracetamol	1.5	0.7
79	3.5	Phenacetin	1.5	0.7
80	3.5	Starch	1.5	0.7
81	4.0	Caffeine	1.0	0.8
82	4.0	Glucose	1.0	0.8
83	4.0	Paracetamol	1.0	0.8
84	4.0	Phenacetin	1.0	0.8
85	4.0	Starch	1.0	0.8
86	4.5	Caffeine	0.5	0.9
87	4.5	Glucose	0.5	0.9
88	4.5	Paracetamol	0.5	0.9
89	4.5	Phenacetin	0.5	0.9
90	4.5	Starch	0.5	0.9

Table 1. Composition of heroin and methamphetamine mixed sample

Sample name	PCA1	PCA2	PCA3	PCA4	PCA5	PCA6	PCA7	PCA8
Heroin 10	-1.73499	1.05294	-0.24404	0.54695	1.23830	1.86434	-0.85063	0.30997
Heroin 20	-1.38742	1.08753	-0.49341	0.63136	1.24009	1.91988	-0.82360	0.01083
Heroin 30	-1.36164	1.31395	-0.46624	0.02419	0.53490	1.55977	0.99659	0.82939
Heroin 40	-1.26595	1.30657	-0.48177	-0.02275	0.57038	1.41293	0.98688	0.84195
Heroin 50	-1.42631	1.11977	-0.43161	-0.06417	0.41947	1.82224	0.55125	0.68340
Methamphetamine 10	-1.82849	0.65761	-0.19163	0.39233	1.02352	2.11987	-1.59872	0.70246
Methamphetamine 20	-1.62927	-0.30444	-0.28452	-0.26524	0.56580	1.76207	-1.45115	0.99943
Methamphetamine 30	-1.61720	-0.11446	-0.20642	-0.04146	0.62600	1.77511	-0.81298	0.14576
Methamphetamine 40	-1.37645	-0.47555	-0.20890	-0.02513	0.67996	1.46555	-0.43467	0.06310
Methamphetamine 50	-1.27684	-0.60591	-0.35037	-0.09153	0.58028	1.56357	-0.93970	0.27797
Sample name	PCA9	PCA10	PCA11	PCA12	PCA13	PCA14	PCA15
Heroin 10	0.66874	0.04875	0.14664	0.23064	0.50656	-0.15218	0.11773
Heroin 20	0.96444	-0.50303	0.21401	0.47814	1.49474	-0.32094	-0.05898
Heroin 30	1.40614	0.64159	0.52790	0.61595	-0.82069	0.88689	-0.50806
Heroin 40	1.35271	0.36899	0.55841	0.43866	-0.34179	0.78488	-0.37276
Heroin 50	1.19463	0.66721	0.21536	1.00858	-0.25017	0.69153	-0.39813
Methamphetamine 10	0.18801	0.11184	-0.28867	-0.68909	-0.52643	-0.59025	0.38801
Methamphetamine 20	-0.12065	0.07006	-0.46789	-0.04990	-0.34572	-0.95516	0.10060
Methamphetamine 30	0.46265	0.17059	-0.36223	-0.22779	-0.59669	-0.41832	-0.08377
Methamphetamine 40	0.58264	0.26791	-0.27522	0.82319	0.05651	-0.76757	-0.06932
Methamphetamine 50	0.32392	0.04199	-0.44118	0.03576	0.18528	-0.67303	0.35902

Table 2. Principal component analysis scores of partial data

Function	Comprehensive classification accuracy	Heroin classification accuracy
RBF-SVM	48.9	97.8
Polynomial-SVM	47.8	95.6
Sigmoid-SVM	35.6	71.2
Linear-SVM	48.9	97.8

Table 3. Accuracy of sample classification under four kernel functions unit: %

Gama	Number of heroin samples	Number of correctly classified samples	Classification accuracy /%
0.01	45	39	86.7
0.05	45	39	86.7
0.10	45	43	95.6
1.00	45	43	95.6
10.00	45	43	95.6
100.00	45	43	95.6

Table 4. Classification effect with polynomial SVM model with different Gamma values

Training sample ratio	Constitution	Additives
50	84.9	96.5
60	80.4	91.5
70	77.8	91.9
80	74.2	85.9
90	71.1	88.4

Table 5. Average classification accuracy under different training ratios unit: %

Wenjie Yan, Wenhui Lu, Jifen Wang. Research on Spectral Recognition of Drug Mixture Based on SVM-MLP Fusion Model[J]. Laser & Optoelectronics Progress, 2021, 58(14): 1404003

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information