Identification of Kidney-Yang Deficiency Syndrome in Osteoporosis Patients Based on Rule Ensemble Method of Bagging Combining LASSO Regression / 中国实验方剂学杂志

ObjectiveTo investigate the identification of kidney Yang deficiency syndrome of patients with osteoporosis（OP）， and to form the clinical syndrome identification rules of traditional Chinese medicine（TCM）. MethodBasic information， etiology， clinical symptoms and other characteristics of 982 OP patients were included， and statistical tests were used to screen the variables associated with kidney Yang deficiency syndrome. Taking the decision tree as the base model， bootstrap aggregation algorithm（Bagging algorithm） was utilized to establish the classification model of kidney Yang deficiency syndrome in OP， generating numerous rules and removing redundancy. Combining least absolute shrinkage and selection operator（LASSO） regression to screen key rules and integrate them to construct an identification model， achieving the identification of kidney Yang deficiency syndrome in OP patients. ResultEighteen key identification rules were screened out， and of these， where 11 rules with regression coefficients>0 correlated positively with the kidney Yang deficiency syndrome， the rule with the highest coefficient was chilliness（present）&feverish sensation over the palm and sole（absent）. The other 7 rules with regression coefficients<0 correlated negatively with the syndrome， the rule with the lowest coefficient was reddish tongue（present）&diarrhea（absent）&deficiency of endowment（absent）. According to the regression coefficients of each key rule， variables with importance>0.2 were ranked as chilliness， reddish tongue， feverish sensation over the palm and sole， cold limbs， clear urine， diarrhea， deficiency of endowment， prolonged illness. The results of the partial dependence analysis of the identification model showed that compared to OP patients without chilliness， those with chilliness（present） had a 0.266 8 higher probability of being identified as having kidney Yang deficiency syndrome， indicating that this variable had the highest impact on identification of the syndrome. Similarly， compared to OP patients without reddish tongue， those with reddish tongue had a 0.141 9 lower probability of being identified as having kidney Yang deficiency syndrome， indicating that this variable had the highest impact on identifying non-kidney Yang deficiency syndrome. The accuracy， sensitivity， specificity and area under receiver operating characteristic curve（AUC） of the established kidney Yang deficiency syndrome identification model in the test set were 0.865 9， 0.853 7， 0.872 0 and 0.931 5， respectively. ConclusionA precise identification model of OP kidney Yang deficiency syndrome is conducted basing on the rule ensemble method of Bagging combining LASSO regression， and the screened key rules can explain the identification process of kidney Yang deficiency syndrome. In this research， according to the regression coefficients of rules， the importance and partial dependence of variables， combined with the thinking of TCM， the influence of patient characteristics on the identification of syndromes is described， so as to reveal the primary and secondary syndromes of identification and assist the clinical identification of kidney Yang deficiency syndrome.

iSucc-PseAAC:Prediction of Lysine Succinylation Modification Sites Based on Ensemble Machine Learning / 中国生物化学与分子生物学报

Lysine succinylation is a novel post-translational modification, which plays an important role in regulating distinct cellular functions control, therefore it is necessary to accurately identify succinylation sites in proteins. As traditional experiments consume material and financial resources, prediction by calculation is an efficient method being proposed recently. In this study, we developed a new prediction method iSucc-PseAAC, which uses a variety of classification algorithms combined with different feature extraction methods. Moreover, it is found that under the feature extraction based on coupled sequence (PseAAC), the classification effect of support vector machine is the best, and it could be combined with ensemble learning to solve the problem of data imbalance. Compared with the existing methods, iSucc-PseAAC has more significance and practicality in predicting lysine succinylation sites.

Complete Box Fusion based on Ensemble Networks for rib fracture detection and localization / 解剖学报

Objective To propose a new rib fracture detection network Rib-Net to automatically and accurately detect and locate rib fracture and address the issue of missed diagnosis of rib fractures. Methods The public data set RibFrac Dataset was used to evaluate the performance of the Rib-Net, and the data set was divided into training set (420 cases), validation set (80 cases), and test set (160 cases). The Rib-Net was composed of the object detection integrated network Ensemble Detection Net (ED-Ne), Complete Box Fusion (CBF) module and the segmentation network 3D Unet. Firstly, Retina Unet, UFRCNN+ and Mask RCNN were integrated to form ED-Net to predict rib fracture candidate boxes. Secondly, a new CBF module was designed to fuse overlapping fracture candidate boxes to generate candidate boxes with accurate positioning and accurate confidence. Finally, Unet was used for rib fracture segmentation to achieve further precise localization of rib fractures. Results On the “MICCAI 2020 RibFrac Challenge: Rib Fracture Detection and Classification challenge”, our proposed Rib-Net’s detection results reached the best performance, and its recall rate, free-response receiver operating characteristic curve(FROC) value and Dice were 92.3%, 0.859 and 0.61, respectively. Conclusion The Rib-Net network can efficiently and accurately detect and locate rib fractures on chest CT images, effectively assisting doctors in making accurate diagnosis.

A heart sound classification method based on complete ensemble empirical modal decomposition with adaptive noise permutation entropy and support vector machine / 生物医学工程学杂志

Heart sound signal is a kind of physiological signal with nonlinear and nonstationary features. In order to improve the accuracy and efficiency of the phonocardiogram (PCG) classification, a new method was proposed by means of support vector machine (SVM) in which the complete ensemble empirical modal decomposition with adaptive noise (CEEMDAN) permutation entropy was as the eigenvector of heart sound signal. Firstly, the PCG was decomposed by CEEMDAN into a number of intrinsic mode functions (IMFs) from high to low frequency. Secondly, the IMFs were sifted according to the correlation coefficient, energy factor and signal-to-noise ratio. Then the instantaneous frequency was extracted by Hilbert transform, and its permutation entropy was constituted into eigenvector. Finally, the accuracy of the method was verified by using a hundred PCG samples selected from the 2016 PhysioNet/CinC Challenge. The results showed that the accuracy rate of the proposed method could reach up to 87%. In comparison with the traditional EMD and EEMD permutation entropy methods, the accuracy rate was increased by 18%-24%, which demonstrates the efficiency of the proposed method.

Epilepsy detection and analysis method for specific patient based on data augmentation and deep learning / 生物医学工程学杂志

In recent years, epileptic seizure detection based on electroencephalogram (EEG) has attracted the widespread attention of the academic. However, it is difficult to collect data from epileptic seizure, and it is easy to cause over fitting phenomenon under the condition of few training data. In order to solve this problem, this paper took the CHB-MIT epilepsy EEG dataset from Boston Children's Hospital as the research object, and applied wavelet transform for data augmentation by setting different wavelet transform scale factors. In addition, by combining deep learning, ensemble learning, transfer learning and other methods, an epilepsy detection method with high accuracy for specific epilepsy patients was proposed under the condition of insufficient learning samples. In test, the wavelet transform scale factors 2, 4 and 8 were set for experimental comparison and verification. When the wavelet scale factor was 8, the average accuracy, average sensitivity and average specificity was 95.47%, 93.89% and 96.48%, respectively. Through comparative experiments with recent relevant literatures, the advantages of the proposed method were verified. Our results might provide reference for the clinical application of epilepsy detection.

Applying Biomod2 for modeling of species suitable habitats:a case study of Paeonia lactiflora in China / 中国中药杂志

Paeonia lactiflora is an important medicinal resource in China. It is of great significance for the protection and cultivation of P. lactiflora resources to find the suitable habitats. The study was based on the information of 98 distribution sites and the data of 20 current environmental factors of wild P. lactiflora in China. According to the correlation and importance of environmental factors, we selected the main environmental factors affecting the potential suitable habitats. Then, BCC-CSM2-MR model was employed to predict the distribution range and center change of potential suitable habitat of wild P. lactiflora in the climate scenarios of SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5 during 2021-2100. The ensemble model combined with GBM, GLM, MaxEnt, and RF showed improved prediction accuracy, with TSS=0.85 and AUC=0.95. Among the 20 environmental factors, annual mean temperature, monthly mean diurnal range of temperature, temperature seasonality, mean temperature of the warmest quarter, precipitation of the wettest month, precipitation seasonality, precipitation of the driest quarter, and elevation were the main factors that affected the suitable habitat distribution of P. lactiflora. At present, the potential suitable habitats of wild P. lactiflora is mainly distributed in Inner Mongolia, Heilongjiang, Jilin, Liaoning, Hebei, Beijing, Shaanxi, Shanxi, Shandong, Gansu, Xinjiang, Tibet, and Ningxia, and concentrated in the northeastern Inner Mongolia, central Heilongjiang, and northern Jilin. Under future climate conditions, the highly sui-table area of wild P. lactiflora will shrink, and the potential suitable habitat will mainly be lost to different degrees. However, in the SSP5-8.5 scenario, the low suitable area of wild P. lactiflora will partially increase in the highlands and mountains in western China including Xinjiang, Tibet, and Qinghai during 2061-2100. The distribution center of wild P. lactiflora migrated first to the northeast and then to the southwest. The total suitable habitats were stable and kept in the high-latitude zones. The prediction of the potential geo-graphical distribution of P. lactiflora is of great significance to the habitat protection and standardized cultivation of this plant in the future.

iSulf_wide-PseAAC: A Protein S-sulfonylation Prediction Method Based on Ensemble Support Vector Machine Approach / 中国生物化学与分子生物学报

Protein S-sulfonylation is a reversible protein post-translational modification (PTM) that plays a vital role in biological growth. It is also related to some diseases. Therefore from the perspective of basic research or drug development, we are faced with a challenging question: where are the S-sulfonylation sites on proteins? In order to solve this problem, we developed a method for site prediction. The working principle of the system is (1) Combine these proteins into isometric amino acids.(2) Use down sampling to balance the training data set;(3) Build a comprehensive prediction system through ensemble support vector machines. In the end, the accuracy rate on the independent tester reached 90. 77%. Compared with the existing methods, all other indicators have been improved, and the improvement effect is obvious. Thus we provided help for the development of bioinformatics. We have established a friendly web server prediction website: http://www.jci-bioinfo.cn/iSulf_Wide-PseAAC, through which the website can make the prediction online, without complicated calculation formulas. At the same time, the mathematical methods used in this article can solve many problems in similar related fields.

Invisible Protein State / 中国生物化学与分子生物学报

Conformation dynamics attributes to the biological functions of active proteins and conformation ensembles. The conformation ensembles include protein stable states (PSS) that can be measured by conventional structural biology approaches and the invisible protein states (IPS) that cannot be measured by conventional structural biology ones. The conformational exchange between IPS and PSS plays an important role in the biological functions of proteins. In this review, we briefly describe the basic properties of IPS and discuss its contribution to the development of the classical molecular recognition mechanism of “keylock hypothesis” and "induced fit hypothesis" into "conformational selection hypothesis". Furthermore, this review compares the advantages and disadvantages of the current structural biology approaches for investigating the conformational properties of IPS. Because of advanced NMR technology, the exploration of the conformational properties of IPS in experimental manner has become feasible. It is expected that the study of IPS and its function will not only help clarify the molecule recognition mechanism of proteins, but also provide a basis for guiding the design of targeted drugs.

Application of multiple empirical kernel mapping ensemble classifier based on self-paced learning in ultrasound-based computer-aided diagnosis for breast cancer / 生物医学工程学杂志

Both feature representation and classifier performance are important factors that determine the performance of computer-aided diagnosis (CAD) systems. In order to improve the performance of ultrasound-based CAD for breast cancers, a novel multiple empirical kernel mapping (MEKM) exclusivity regularized machine (ERM) ensemble classifier algorithm based on self-paced learning (SPL) is proposed, which simultaneously promotes the performance of both feature representation and the classifier. The proposed algorithm first generates multiple groups of features by MEKM to enhance the ability of feature representation, which also work as the kernel transform in multiple support vector machines embedded in ERM. The SPL strategy is then adopted to adaptively select samples from easy to hard so as to gradually train the ERM classifier model with improved performance. This algorithm is verified on a B-mode ultrasound dataset and an elastography ultrasound dataset, respectively. The results show that the classification accuracy, sensitivity and specificity on B-mode ultrasound are (86.36±6.45)%, (88.15±7.12)%, and (84.52±9.38)%, respectively, and the classification accuracy, sensitivity and specificity on elastography ultrasound are (85.97±3.75)%, (85.93±6.09)%, and (86.03±5.88)%, respectively. It indicates that the proposed algorithm can effectively improve the performance of ultrasound-based CAD for breast cancers with the potential for application.

Psychosis speech recognition algorithm based on deep embedded sparse stacked autoencoder and manifold ensemble / 生物医学工程学杂志

Speech feature learning is the core and key of speech recognition method for mental illness. Deep feature learning can automatically extract speech features, but it is limited by the problem of small samples. Traditional feature extraction (original features) can avoid the impact of small samples, but it relies heavily on experience and is poorly adaptive. To solve this problem, this paper proposes a deep embedded hybrid feature sparse stack autoencoder manifold ensemble algorithm. Firstly, based on the prior knowledge, the psychotic speech features are extracted, and the original features are constructed. Secondly, the original features are embedded in the sparse stack autoencoder (deep network), and the output of the hidden layer is filtered to enhance the complementarity between the deep features and the original features. Third, the L1 regularization feature selection mechanism is designed to compress the dimensions of the mixed feature set composed of deep features and original features. Finally, a weighted local preserving projection algorithm and an ensemble learning mechanism are designed, and a manifold projection classifier ensemble model is constructed, which further improves the classification stability of feature fusion under small samples. In addition, this paper designs a medium-to-large-scale psychotic speech collection program for the first time, collects and constructs a large-scale Chinese psychotic speech database for the verification of psychotic speech recognition algorithms. The experimental results show that the main innovation of the algorithm is effective, and the classification accuracy is better than other representative algorithms, and the maximum improvement is 3.3%. In conclusion, this paper proposes a new method of psychotic speech recognition based on embedded mixed sparse stack autoencoder and manifold ensemble, which effectively improves the recognition rate of psychotic speech.

Research on automatic removal of ocular artifacts from single channel electroencephalogram signals based on wavelet transform and ensemble empirical mode decomposition / 生物医学工程学杂志

The brain-computer interface (BCI) systems used in practical applications require as few electroencephalogram (EEG) acquisition channels as possible. However, when it is reduced to one channel, it is difficult to remove the electrooculogram (EOG) artifacts. Therefore, this paper proposed an EOG artifact removal algorithm based on wavelet transform and ensemble empirical mode decomposition. Firstly, the single channel EEG signal is subjected to wavelet transform, and the wavelet components which involve EOG artifact are decomposed by ensemble empirical mode decomposition. Then the predefined autocorrelation coefficient threshold is used to automatically select and remove the intrinsic modal functions which mainly composed of EOG components. And finally the 'clean' EEG signal is reconstructed. The comparative experiments on the simulation data and the real data show that the algorithm proposed in this paper solves the problem of automatic removal of EOG artifacts in single-channel EEG signals. It can effectively remove the EOG artifacts when causes less EEG distortion and has less algorithm complexity at the same time. It helps to promote the BCI technology out of the laboratory and toward commercial application.

Diabetes Prediction Using Ensemble Classifier

Diabetes is one of the impactful diseases that affect humans’ health rigorously. Early diagnosis of diabetes will assist health caresystems to decide and act according to counter measures. This paper focuses on obtaining an automated tool that will predictdiabetic tendency of a patient. The system proposed by this paper contains two ensemble classifiers- Voting ensemble classifierand Stacking Ensemble classifier. Both of these methods exhibits better results while compared to other classifiers. Stackingensemble classifier even performs better than voting ensemble classifier with an accuracy of 79.87%.

Research on heart rate extraction algorithm in motion state based on normalized least mean square combining ensemble empirical mode decomposition / 生物医学工程学杂志

In order to eliminate the influence of motion artifacts, high-frequency noise and baseline drift on photoplethysmographic (PPG), and to obtain the accurate value of heart rate in motion state, this paper proposed a de-noising method of PPG signal based on normalized least mean square (NLMS) adaptive filtering combining ensemble empirical mode decomposition(EEMD). Firstly, the PPG signal containing noise is passed through an adaptive filter with a 3-axis acceleration sensor as a reference signal to filter out motion artifacts. Secondly, the PPG signal is decomposed by EEMD to obtain a series of intrinsic modal function (IMF) according to the frequency from high to low. The threshold range of the signal is judged by the permutation entropy (PE) criterion, thereby filtering out the high frequency noise and the baseline drift. The experimental results show that the Pearson correlation coefficient between the calculated heart rate of PPG signal and the standard heart rate based on electrocardiogram (ECG) signal is 0.731 and the average absolute error percentage is 6.10% under different motion states, which indicates that the method can accurately calculate the heart rate in moving state and is beneficial to the physiological monitoring under the state of human motion.

A spike denoising method combined principal component analysis with wavelet and ensemble empirical mode decomposition / 生物医学工程学杂志

Spike recorded by multi-channel microelectrode array is very weak and susceptible to interference, whose noisy characteristic affects the accuracy of spike detection. Aiming at the independent white noise, correlation noise and colored noise in the process of spike detection, combining principal component analysis (PCA), wavelet analysis and adaptive time-frequency analysis, a new denoising method (PCWE) that combines PCA-wavelet (PCAW) and ensemble empirical mode decomposition is proposed. Firstly, the principal component was extracted and removed as correlation noise using PCA. Then the wavelet-threshold method was used to remove the independent white noise. Finally, EEMD was used to decompose the noise into the intrinsic modal function of each layer and remove the colored noise. The simulation results showed that PCWE can increase the signal-to-noise ratio by about 2.67 dB and decrease the standard deviation by about 0.4 μV, which apparently improved the accuracy of spike detection. The results of measured data showed that PCWE can increase the signal-to-noise ratio by about 1.33 dB and reduce the standard deviation by about 18.33 μV, which showed its good denoising performance. The results of this study suggests that PCWE can improve the reliability of spike signal and provide an accurate and effective spike denoising new method for the encoding and decoding of neural signal.

Single-modal neuroimaging computer aided diagnosis for schizophrenia based on ensemble learning using privileged information / 生物医学工程学杂志

Neuroimaging technologies have been applied to the diagnosis of schizophrenia. In order to improve the performance of the single-modal neuroimaging-based computer-aided diagnosis (CAD) for schizophrenia, an ensemble learning algorithm based on learning using privileged information (LUPI) was proposed in this work. Specifically, the extreme learning machine based auto-encoder (ELM-AE) was first adopted to learn new feature representation for the single-modal neuroimaging data. Random project algorithm was then performed on the learned high-dimensional features to generate several new feature subspaces. After that, multiple feature pairs were built among these subspaces to work as source domain and target domain, respectively, which were used to train multiple support vector machine plus (SVM+) classifier. Finally, a strong classifier is learned by combining these SVM+ classifiers for classification. The proposed algorithm was evaluated on a public schizophrenia neuroimaging dataset, including the data of structural magnetic resonance imaging (sMRI) and functional MRI (fMRI). The results showed that the proposed algorithm achieved the best diagnosis performance. In particular, the classification accuracy, sensitivity and specificity of the proposed algorithm were 72.12% ± 8.20%, 73.50% ± 15.44% and 70.93% ± 12.93%, respectively, on the sMRI data, and it also achieved the classification accuracy of 72.33% ± 8.95%, sensitivity of 68.50% ± 16.58% and specificity of 75.73% ± 16.10% on the fMRI data. The proposed algorithm overcomes the problem that the traditional LUPI methods need the additional privileged information modality as source domain. It can be directly applied to the single-modal data for classification, and also can improve the classification performance. Therefore, it suggests that the proposed algorithm will have wider applications.

On the necessity of an integrative approach to understand protein structural dynamics / 浙江大学学报（英文版）（B辑：生物医学和生物技术）

Proteins are dynamic, fluctuating between multiple conformational states. Protein dynamics, spanning orders of magnitude in time and space, allow proteins to perform specific functions. Moreover, under certain conditions, proteins can morph into a different set of conformations. Thus, a complete understanding of protein structural dynamics can provide mechanistic insights into protein function. Here, we review the latest developments in methods used to determine protein ensemble structures and to characterize protein dynamics. Techniques including X-ray crystallography, cryogenic electron microscopy, and small angle scattering can provide structural information on specific conformational states or on the averaged shape of the protein, whereas techniques including nuclear magnetic resonance, fluorescence resonance energy transfer (FRET), and chemical cross-linking coupled with mass spectrometry provide information on the fluctuation of the distances between protein domains, residues, and atoms for the multiple conformational states of the protein. In particular, FRET measurements at the single-molecule level allow rapid resolution of protein conformational states, where information is otherwise obscured in bulk measurements. Taken together, the different techniques complement each other and their integrated use can offer a clear picture of protein structure and dynamics.

Processing of impedance cardiogram differential for non-invasive cardiac function detection / 生物医学工程学杂志

The precise recognition of feature points of impedance cardiogram (ICG) is the precondition of calculating hemodynamic parameters based on thoracic bioimpedance. To improve the accuracy of detecting feature points of ICG signals, a new method was proposed to de-noise ICG signal based on the adaptive ensemble empirical mode decomposition and wavelet threshold firstly, and then on the basis of adaptive ensemble empirical mode decomposition, we combined difference and adaptive segmentation to detect the feature points, A, B, C and X, in ICG signal. We selected randomly 30 ICG signals in different forms from diverse cardiac patients to examine the accuracy of the proposed approach and the accuracy rate of the proposed algorithm is 99.72%. The improved accuracy rate of feature detection can help to get more accurate cardiac hemodynamic parameters on the basis of thoracic bioimpedance.

On the necessity of an integrative approach to understand protein structural dynamics / 浙江大学学报（英文版）（B辑：生物医学和生物技术）

Proteins are dynamic, fluctuating between multiple conformational states. Protein dynamics, spanning orders of magnitude in time and space, allow proteins to perform specific functions. Moreover, under certain conditions, proteins can morph into a different set of conformations. Thus, a complete understanding of protein structural dynamics can provide mechanistic insights into protein function. Here, we review the latest developments in methods used to determine protein ensemble structures and to characterize protein dynamics. Techniques including X-ray crystallography, cryogenic electron microscopy, and small angle scattering can provide structural information on specific conformational states or on the averaged shape of the protein, whereas techniques including nuclear magnetic resonance, fluorescence resonance energy transfer (FRET), and chemical cross-linking coupled with mass spectrometry provide information on the fluctuation of the distances between protein domains, residues, and atoms for the multiple conformational states of the protein. In particular, FRET measurements at the single-molecule level allow rapid resolution of protein conformational states, where information is otherwise obscured in bulk measurements. Taken together, the different techniques complement each other and their integrated use can offer a clear picture of protein structure and dynamics.

A partition bagging ensemble learning algorithm for Parkinson's speech data mining / 生物医学工程学杂志

Methods for achieving diagnosis of Parkinson's disease (PD) based on speech data mining have been proven effective in recent years. However, due to factors such as the degree of disease of the data collection subjects and the collection equipment and environment, there are different categories of sample aliasing in the sample space of the acquired data set. Samples in the aliased area are difficult to be identified effectively, which seriously affects the classification accuracy of the algorithm. In order to solve this problem, a partition bagging ensemble learning is proposed in this article, which measures the aliasing degree of the sample by designing the the ratio of sample centroid distance metrics and divides the training set into multiple subsets. And then the method of transfer training of misclassified samples is used to adjust the results of subset partitioning. Finally, the optimized weights of each sub-classifier are used to integrate the test results. The experimental results show that the classification accuracy of the proposed method is significantly improved on two public datasets and the increasement of mean accuracy is up to 25.44%. This method not only effectively improves the classification accuracy of PD speech dataset, but also increases the sample utilization rate, providing a new idea for the diagnosis of PD.

Early prognosis of Alzheimer's disease based on convolutional neural networks and ensemble learning / 生物医学工程学杂志

Alzheimer's disease (AD) is a typical neurodegenerative disease, which is clinically manifested as amnesia, loss of language ability and self-care ability, and so on. So far, the cause of the disease has still been unclear and the course of the disease is irreversible, and there has been no cure for the disease yet. Hence, early prognosis of AD is important for the development of new drugs and measures to slow the progression of the disease. Mild cognitive impairment (MCI) is a state between AD and healthy controls (HC). Studies have shown that patients with MCI are more likely to develop AD than those without MCI. Therefore, accurate screening of MCI patients has become one of the research hotspots of early prognosis of AD. With the rapid development of neuroimaging techniques and deep learning, more and more researchers employ deep learning methods to analyze brain neuroimaging images, such as magnetic resonance imaging (MRI), for early prognosis of AD. Hence, in this paper, a three-dimensional multi-slice classifiers ensemble based on convolutional neural network (CNN) and ensemble learning for early prognosis of AD has been proposed. Compared with the CNN classification model based on a single slice, the proposed classifiers ensemble based on multiple two-dimensional slices from three dimensions could use more effective information contained in MRI to improve classification accuracy and stability in a parallel computing mode.