Feature extraction for identification of drug body packing based on nonnegative matrix factorization

Fast and accurate determination of organophosphate pesticides using surface-enhanced Raman scattering and chemometrics

“…36 The accumulating contribution rate for PCA was up to 95%. NMF 32 is a top-down generative algorithm that optimizes its internal representation to minimize the reconstruction error between the input and the reconstructed output. In addition, its weighted least-squares problem resolution can prevent the occurrence of negative factors and avoid contradicting the physical reality, which is important because the negative factors for the actual problems cannot be explained.…”

“…26 Finite difference (FD), 27 multiplicative scatter correction (MSC) 28 and standard normal variate variance (SNV) transformation 29 were then adopted to decrease the noise. Principal components analysis (PCA) 30,31 and non-negative matrix factorization (NMF) 32 were used to extract the main information and reduce the dimension of the spectral data. Various data were then used to develop the classication models by SVM and RF.…”

Prediction of protein phenotype based on protein interaction network by coupling genetic algorithm and K-nearest neighbor algorithm

“…In recent years, the K-nearest neighbor algorithm has been widely used in the eld of bioinformatics. [15][16][17][18] However, when using the K-nearest neighbor algorithm for predicting the various properties of protein, two problems must be taken into account, namely the choice of the parameter K and feature subset because they have important impacts on the performance of the model. Therefore, the parameter K and feature subset must be optimized to obtain the model with the best performance.…”

Marginal Fisher Analysis-based feature extraction for identification of drug and explosive concealed by body packing

“…Among various techniques, such as the uorescence of dual-emission quantum dot hybrids, 1 fast neutron scattering, 2 CT, 3 terahertz imaging 4 and homogeneous phase protein-based assays, 5 energy dispersive X-ray diffraction (EDXRD) for detection within complex backgrounds has been proved to be a promising method due to its non-destructiveness, high selectivity and high efficiency. [6][7][8][9][10][11][12][13][14][15][16][17][18][19] This technology is based on measurement of the energy-dispersive low-angle scattering of photons from a polychromatic incident beam. For EDXRD testing, the EDXRD spectrum is a diffraction prole from the scattering media.…”

“…As a method of feature extraction, principal component analysis (PCA) has been used in the analysis of EDXRD spectra. 6,9,[15][16][17][18][19] According to the mathematical principle of PCA, it is an optimal linear dimensional reduction technique in the mean-square sense and seeks projection directions with maximal variances. PCA searches for directions efficient for representation, but fails to make use of class information.…”