Training Sample Set Research Articles

Comparison and optimization of rapid and nondestructive qualitative analysis models based on fourier transform infrared spectroscopy information of drugs

It is a key issue in the field of drug prevention and control to realize the rapid and nondestructive qualitative analysis of drugs. In this study, spectral data of 196 pieces of drug samples based on actual cases which are three types in total, were sorted out. The Bayesian discriminant analysis (BDA) models were constructed based on full spectral, fingerprint spectral and functional group spectral data sets were extracted by principal component analysis (PCA). The classification results for different types of data sets were compared, and the optimal spectral region were selected. In the BDA, the model based on the data of the fingerprint spectra the highest classification accuracy, and the classification accuracy of the sample training set based on three types of drugs was 0.837. The identification models based on data sets of fingerprint spectra were established by multilayer perceptron (MLP) and support vector machines (SVM), as well as compare the model parameters of SVM models based on different kernel functions. According to the results, it shows that in the three types of drugs recognition and classification process, the SVM model based on linear kernel functions has achieved the best classification effect, with the overall classification accuracy of the sample training set reaches 0.849, and the model can fully classify drugs of the same type. The classification results were ideal for the SVM model based on linear kernel functions for fingerprint spectral data set. The results of this study demonstrate the potential of Fourier transform infrared spectroscopy in combination with the chemometrics methods of SVM as a new method for the identification of different types and different kinds of drugs of the same type.

Static Voltage Stability Assessment Using a Random UnderSampling Bagging BP Method

The increase in demand and generator reaching reactive power limits may operate the power system in stressed conditions leading to voltage instability. Thus, the voltage stability assessment is essential for estimating the loadability margin of the power system. The grid operators urgently need a voltage stability assessment (VSA) method with high accuracy, fast response speed, and good scalability. The static VSA problem is defined as a regression problem. Moreover, an artificial neural network is constructed for online assessment of the regression problem. Firstly, the training sample set is obtained through scene simulation, power flow calculation, and local voltage stability index calculation; then, the class imbalance problem of the training samples is solved by the random under-sampling bagging (RUSBagging) method. Then, the mapping relationship between each feature and voltage stability is obtained by an artificial neural network. Finally, taking the modified IEEE39 node system as an example, by setting up four groups of methods for comparison, it is verified that the proposed method has a relatively ideal modeling speed and high accuracy, and can meet the requirements of power system voltage stability assessment.

Identification and immune characteristics of molecular subtypes related to fatty acid metabolism in idiopathic pulmonary fibrosis.

BackgroundAlthough fatty acid metabolism has been confirmed to be involved in the pathological process of idiopathic pulmonary fibrosis (IPF), systematic analyses on the immune process mediated by fatty acid metabolism-related genes (FAMRGs) in IPF remain lacking.MethodsThe gene expression data of 315 patients with IPF were obtained from Gene Expression Omnibus database and were divided into the training and verification sets. The core FAMRGs of the training set were identified through weighted gene co-expression network analysis. Then, the fatty acid metabolism-related subtypes in IPF were identified on the basis of k-means unsupervised clustering. The scores of fatty acid metabolism and the expression of the fibrosis biomarkers in different subtypes were compared, and functional enrichment analysis was carried out on the differentially expressed genes between subtypes. A random forest model was used to select important FAMRGs as diagnostic markers for distinguishing between subtypes, and a line chart model was constructed and verified by using other datasets and rat models with different degrees of pulmonary fibrosis. The difference in immune cell infiltration among subtypes was evaluated with CIBERSORT, and the correlation between core diagnostic markers and immune cells were analyzed.ResultsTwenty-four core FAMRGs were differentially expressed between the training set and normal samples, and IPF was divided into two subtypes. Significant differences were observed between the two subtypes in biological processes, such as linoleic acid metabolism, cilium movement, and natural killer (NK) cell activation. The subtype with high fatty acid metabolism had more severe pulmonary fibrosis than the other subtype. A reliable construction line chart model based on six diagnostic markers was constructed, and ABCA3 and CYP24A1 were identified as core diagnostic markers. Significant differences in immune cell infiltration were found between the two subtypes, and ABCA3 and CYP24A1 were closely related to NK cells.ConclusionFatty acid metabolism and the immune process that it mediates play an important role in the occurrence and development of IPF. The analysis of the role of FAMRGs in IPF may provide a new potential therapeutic target for IPF.

Prediction of Total-Body and Partial-Body Exposures to Radiation Using Plasma Proteomic Expression Profiles.

There is a need to identify new biomarkers of radiation exposure for not only systemic total-body irradiation (TBI) but also to characterize partial-body irradiation and organ specific radiation injury. In the current study, we sought to develop novel biodosimetry models of radiation exposure using TBI and organ specific partial-body irradiation to only the brain, lung or gut using a multivariate proteomics approach. Subset panels of significantly altered proteins were selected to build predictive models of radiation exposure in a variety of sample cohort configurations relevant to practical field application of biodosimetry diagnostics during future radiological or nuclear event scenarios. Female C57BL/6 mice, 8-15 weeks old, received a single total-body or partial-body dose of 2 or 8 Gy TBI or 2 or 8 Gy to only the lung or gut, or 2, 8 or 16 Gy to only the brain using a Pantak X-ray source. Plasma was collected by cardiac puncture at days 1, 3 and 7 postirradiation for total-body exposures and only the lung and brain exposures, and at days 3, 7 and 14 postirradiation for gut exposures. Plasma was then screened using the aptamer-based SOMAscan proteomic assay technology, for changes in expression of 1,310 protein analytes. A subset panel of protein biomarkers which demonstrated significant changes (P < 0.01) in expression after irradiation were used to build predictive models of radiation exposure using different sample cohorts. Model 1 compared controls vs. all pooled irradiated samples, which included TBI and all organ specific partial irradiation. Model 2 compared controls vs. TBI vs. partial irradiation (with all organ specific partial exposure pooled within the partial-irradiated group), and model 3 compared controls vs. each individual organ specific partial-body exposure separately (brain, gut and lung). Detectable values were obtained for all 1,310 proteins included in the SOMAscan assay for all samples. Each model algorithm built using a unique sample cohort was validated with a training set of samples and tested with a separate new sample series. Overall predictive accuracies of 89%, 78% and 55% resulted for models 1-3, respectively, representing novel predictive panels of radiation responsive proteomic biomarkers. Though relatively high overall predictive accuracies were achieved for models 1 and 2, all three models showed limited accuracy at differentiating between the controls and partial-irradiated body samples. In our study we were able to identify novel panels of radiation responsive proteins useful for predicting radiation exposure and to create predictive models of partial-body exposure including organ specific radiation exposures. This proof-of-concept study also illustrates the inherent physiological limitations of distinguishing between small-body exposures and the unirradiated using proteomic biomarkers of radiation exposure. As use of biodosimetry diagnostics in future mass casualty settings will be complicated by the heterogeneity of partial-body exposure received in the field, further work remains in adapting these diagnostic tools for practical use.

A novel prognostic 6-gene signature for osteoporosis.

IntroductionThe incidence of osteoporosis (OP) keeps increasing due to global aging of the population. Therefore, identifying the diagnostic and prognostic biomarkers of OP is of great significance.MethodsmRNA data from OP and non-OP samples were obtained from GEO database, which were divided into training set (GSE35959) and testing sets (GSE7158, GSE62402, GSE7429 and GSE56815). Gene modules most significantly related to OP were revealed using weighted gene co-expression network analysis (WGCNA) and differentially expressed genes (DEGs) between OP and normal samples in training set were identified using limma R package. Thereafter, above two gene sets were intersected to obtain the genes potentially related to OP. Protein-protein interaction (PPI) pairs were screened by STRING database and visualized using Cytoscape, while the plug-in cytoHubba was used to screen hub genes by determining their topological parameters. Afterwards, a diagnostic model was constructed using those hub genes, whose creditability was further evaluated by testing sets.ResultsThe results of WGCNA analysis found the Black module was most significantly related to OP, which included altogether 1286 genes. Meanwhile, 2771 DEGs were discovered between OP patients and the normal controls. After taking the intersection, 479 genes were identified potentially correlated with the development of OP. Subsequently, six hub genes were discovered through PPI network construction and node topological analysis. Finally, we constructed a support vector machine model based on these six genes, which can accurately classified training and testing set samples into OP and normal groups.ConclusionOur current study constructed a six hub genes-based diagnostic model for OP. Our findings may shed some light on the research of the early diagnosis for OP and had certain practical significance.

Re-Engineered Word Embeddings for Improved Document-Level Sentiment Analysis

In this paper, a novel re-engineering mechanism for the generation of word embeddings is proposed for document-level sentiment analysis. Current approaches to sentiment analysis often integrate feature engineering with classification, without optimizing the feature vectors explicitly. Engineering feature vectors to match the data between the training set and query sample as proposed in this paper could be a promising way for boosting the classification performance in machine learning applications. The proposed mechanism is designed to re-engineer the feature components from a set of embedding vectors for greatly increased between-class separation, hence better leveraging the informative content of the documents. The proposed mechanism was evaluated using four public benchmarking datasets for both two-way and five-way semantic classifications. The resulting embeddings have demonstrated substantially improved performance for a range of sentiment analysis tasks. Tests using all the four datasets achieved by far the best classification results compared with the state-of-the-art.

Construction of Molecular Subtype and Prognosis Prediction Model of Osteosarcoma Based on Aging-Related Genes.

Background Osteosarcoma (OS) is a rare form of malignant bone cancer that is usually detected in young adults and adolescents. This disease shows a poor prognosis owing to its metastatic status and resistance to chemotherapy. Hence, it is necessary to design a risk model that can successfully forecast the OS prognosis in patients. Methods The researchers retrieved the RNA sequencing data and follow-up clinical data related to OS patients from the TARGET and GEO databases, respectively. The coxph function in R software was used for carrying out the Univariate Cox regression analysis for deriving the aging-based genes related sto the OS prognosis. The researchers conducted consistency clustering using the ConcensusClusterPlus R package. The R software package ESTIMATE, MCPcounter, and GSVA packages were used for assessing the immune scores of various subtypes using the ssGSEA technique, respectively. The Univariate Cox and Lasso regression analyses were used for screening and developing a risk model. The ROC curves were constructed, using the pROC package. The performance of their developed risk model and designed survival curve was conducted, with the help of the Survminer package. Results The OS patients were classified into 2 categories, as per the aging-related genes. The results revealed that the Cluster 1 patients showed a better prognosis than the Cluster 2 patients. Both clusters showed different immune microenvironments. Additional screening of the prognosis-associated genes revealed the presence of 5 genes, i.e., ERCC4, GPX4, EPS8, TERT, and STAT5A, and these data were used for developing the risk model. This risk model categorized the training set samples into the high- and low-risk groups. The patients classified into the high-risk group showed a poor OS prognosis compared to the low-risk patients. The researchers verified the reliability and robustness of the designed 5-gene signature using the internal and external datasets. This risk model was able to effectively predict the prognosis even in the samples having differing clinical features. Compared with other models, the 5- gene model performs better in predicting the risk of osteosarcoma. Conclusion The 5-gene signature developed by the researchers in this study could be effectively used for forecasting the OS prognosis in patients.

A novel dynamic radius support vector data description based fault diagnosis method for proton exchange membrane fuel cell systems

Timely fault detection is critical to improving the reliability and durability of the proton exchange membrane fuel cell (PEMFC) system. This paper proposes a novel fault diagnosis method, dynamic radius support vector data description (DR-SVDD), to efficiently identify the PEMFC system's faults. Compared to the classic support vector data description (SVDD) and improved SVDDs, this method considers both the SVDD hypersphere radius information and the distribution characteristics of the training set samples to obtain a more accurate and adequate description of the sample data. The cell voltages and the pressure drops at the cathode and anode obtained experimentally under various fault conditions are chosen as the feature variables for the PEMFC fault diagnosis. The comparative results show that the proposed DR-SVDD strategy performs well in fault class identification for a PEMFC system.

Research on robust control of continuous rotary electro-hydraulic servo motor based on AdaBoost prediction

A robust control strategy based on Adaboost prediction was proposed for the effects of dynamic uncertainty, parameter perturbation and nonlinear factors such as friction and leakage on the robustness and low-speed performance of the electro-hydraulic servo system. Based on the establishment of the Pol-Ind friction model, according to the mathematical model of the continuous rotary electro-hydraulic servo motor, the structural uncertainty model of the system under parameter perturbation and external perturbation and the generalized state equation of the system are established. The robust controller is designed according to the [Formula: see text] control theory. The system feedback mechanism is designed by using the AdaBoost based on the RBF neural network algorithm. The real-time input and output of the system are used as the training sample set for repeated training, and the multiple weak neural network predictors obtained by training are used to form a strong predictor to calculate the forecast output and error of the electro-hydraulic servo system, and the robust controller was used for real-time control. Simulation and experiment show that: compared with the wavelet neural network control, the robust controller based on AdaBoost prediction has a good tracking performance to the slope signal with an input signal of 0.001°/s, and has a high response speed to sinusoidal signal. Under the condition of satisfying the requirement of double 10 index, the maximum response frequency reaches 15 Hz. It can effectively improve the low-speed performance and anti-interference ability of the electro-hydraulic servo system, greatly improve the robustness of the system and expand the frequency band of the system.

Impact of benign sample size on binary classification accuracy

Recently, there has been a significant increase in malware attacks and malicious traffic. Consequently, several machine learning-based detection models have been developed to detect them. However, the detection accuracy of these models is currently evaluated using different methodologies and datasets, with some studies overstating high detection rates. The lack of a common testing approach coupled with the limited datasets used for the experiments make it challenging to compare the performances of these models to identify those that provide superior detection accuracy. A few studies have focused on benign samples and their effects on detection accuracy. The datasets used in the experiments generally consist of benign and malicious samples; hence, binary classification is used in the machine learning models. In the binary classification task, the size of a benign sample affects the classification accuracy of malicious samples, that is, it can either improve or degrade detection accuracy. In this study, we propose a novel metric for evaluating accuracy degradation by increasing benign sample size. We mainly used the FFRI dataset, which consists of 11,243 malware samples and 250,000 benign samples, and evaluated the classification accuracy with extracted strings from the malware. In addition, we obtained other malware samples that we used as supplementary to the main dataset. We increased the number of benign samples for testing by tenfold, while maintaining the malicious sample and benign training sample sizes, which resulted in a decrease of 0.293 in the F1 score. Furthermore, we confirmed that using a sufficiently sized benign training sample set mitigates accuracy degradation. Our metric can be beneficial for evaluating the benign sample size needed in binary classification and comparing accuracy.

Naive Bayesian aided antenna selection for indoor NOMA GSSK VLC network

This paper investigates the light-emitting diodes (LEDs) combinations selection problem in indoor non-orthogonal multiple access (NOMA) enabled generalized space shift keying (GSSK) multi-user visible light communications (VLC) network. To address the high computational complexity in the optimal LEDs combination selection problem, a naive Bayes (NB) aided LEDs combination selection algorithm is proposed to alleviate the high complexity while guaranteeing the system performance. Firstly, the LEDs combination selection process is modeled as a multi-classification problem; Then, by setting the key performance indicator (KPI) as the maximal sum-rate of multiple users, the generated training samples set is utilized to train a classification model by the NB approach; Finally, the classification model is applied to the LEDs combination selection of new users. Simulation results demonstrate that, compared with the optimal LEDs combination selection algorithm, the NB aided selection scheme proposed in this paper can effectively reduce the computational complexity and achieve a better balance between algorithm complexity and network transmission performance.

Automated Detection of Retrogressive Thaw Slumps in the High Arctic Using High-Resolution Satellite Imagery

Retrogressive thaw slumps (RTS) are considered one of the most dynamic permafrost disturbance features in the Arctic. Sub-meter resolution multispectral imagery acquired by very high spatial resolution (VHSR) commercial satellite sensors offer unique capacities in capturing the morphological dynamics of RTSs. The central goal of this study is to develop a deep learning convolutional neural net (CNN) model (a UNet-based workflow) to automatically detect and characterize RTSs from VHSR imagery. We aimed to understand: (1) the optimal combination of input image tile size (array size) and the CNN network input size (resizing factor/spatial resolution) and (2) the interoperability of the trained UNet models across heterogeneous study sites based on a limited set of training samples. Hand annotation of RTS samples, CNN model training and testing, and interoperability analyses were based on two study areas from high-Arctic Canada: (1) Banks Island and (2) Axel Heiberg Island and Ellesmere Island. Our experimental results revealed the potential impact of image tile size and the resizing factor on the detection accuracies of the UNet model. The results from the model transferability analysis elucidate the effects on the UNet model due the variability (e.g., shape, color, and texture) associated with the RTS training samples. Overall, study findings highlight several key factors that we should consider when operationalizing CNN-based RTS mapping over large geographical extents.

Style-based quantum generative adversarial networks for Monte Carlo events

We propose and assess an alternative quantum generator architecture in the context of generative adversarial learning for Monte Carlo event generation, used to simulate particle physics processes at the Large Hadron Collider (LHC). We validate this methodology by implementing the quantum network on artificial data generated from known underlying distributions. The network is then applied to Monte Carlo-generated datasets of specific LHC scattering processes. The new quantum generator architecture leads to a generalization of the state-of-the-art implementations, achieving smaller Kullback-Leibler divergences even with shallow-depth networks. Moreover, the quantum generator successfully learns the underlying distribution functions even if trained with small training sample sets; this is particularly interesting for data augmentation applications. We deploy this novel methodology on two different quantum hardware architectures, trapped-ion and superconducting technologies, to test its hardware-independent viability.

Genetic Programming for High-Level Feature Learning in Crop Classification

Information on crop spatial distribution is essential for agricultural monitoring and food security. Classification with remote-sensing time series images is an effective way to obtain crop distribution maps across time and space. Optimal features are the precondition for crop classification and are critical to the accuracy of crop maps. Although several approaches are available for extracting spectral, temporal, and phenological features for crop identification, these methods depend heavily on domain knowledge and human experiences, adding uncertainty to the final crop classification. This study proposed a novel Genetic Programming (GP) approach to learning high-level features from time series images for crop classification to address this issue. We developed a new representation of GP to extend the GP tree’s width and depth to dynamically generate either fixed or flexible informative features without requiring domain knowledge. This new GP approach was wrapped with four classifiers, i.e., K-Nearest Neighbor (KNN), Decision Tree (DT), Naive Bayes (NB), and Support Vector Machine (SVM), and was then used for crop classification based on MODIS time series data in Heilongjiang Province, China. The performance of the GP features was compared with the traditional features of vegetation indices (VIs) and the advanced feature learning method Multilayer Perceptron (MLP) to show GP effectiveness. The experiments indicated that high-level features learned by GP improved the classification accuracies, and the accuracies were higher than those using VIs and MLP. GP was more robust and stable for diverse classifiers, different feature numbers, and various training sample sets compared with classification using VI features and the classifier MLP. The proposed GP approach automatically selects valuable features from the original data and uses them to construct high-level features simultaneously. The learned features are explainable, unlike those of a black-box deep learning model. This study demonstrated the outstanding performance of GP for feature learning in crop classification. GP has the potential of becoming a mainstream method to solve complex remote sensing tasks, such as feature transfer learning, image classification, and change detection.

Independent verification of circulating miRNA as diagnostic biomarkers for urothelial carcinoma.

Urothelial carcinoma (UC) is an umbrella term for bladder cancers (BCa) and upper‐tract urothelial carcinoma (UTUC), with BCa and UTUC sometimes detected concomitantly. The methods of detection for UC are often inaccurate or highly invasive, and, therefore, are thought to be unsatisfactory. Previously, we reported seven serum miRNAs as diagnostic markers for BCa. Here, we re‐evaluated potential diagnostic miRNAs in different institutions. We prospectively analyzed serum samples obtained from 126 UC patients (BCa: 106 samples; UTUC: 14 samples; UTUC with BCa: six samples) and 50 noncancer controls by microarray analysis. We randomly assigned these samples into a training or a validation set. Biomarker candidate miRNAs were selected based on cross‐validation scores in the training set of samples, with diagnostic power confirmed in the validation set. Among the diagnostic miRNAs identified in this way, miR‐1343‐5p and miR‐6087 had been identified as potential diagnostic miRNAs in our previous study. In addition, we evaluated the association between the serum levels of identified miRNAs and the presence of UC risk conditions. The expression levels of several miRNAs correlate with the risk factors in participants without UC, which may be explained by the presence of a microscopic tumor or a precancerous lesion. In conclusion, we identified two robust miRNA diagnostic markers for UC detection. Further functional analysis is required to elucidate the mechanism by which alterations in the expression of these miRNAs occur.

Above-Ground Biomass Prediction for Croplands at a Sub-Meter Resolution Using UAV–LiDAR and Machine Learning Methods

Current endeavors to enhance the accuracy of in situ above-ground biomass (AGB) prediction for croplands rely on close-range monitoring surveys that use unstaffed aerial vehicles (UAVs) and mounted sensors. In precision agriculture, light detection and ranging (LiDAR) technologies are currently used to monitor crop growth, plant phenotyping, and biomass dynamics at the ecosystem scale. In this study, we utilized a UAV–LiDAR sensor to monitor two crop fields and a set of machine learning (ML) methods to predict real-time AGB over two consecutive years in the region of Mid-Jutland, Denmark. During each crop growing period, UAV surveys were conducted in parallel with AGB destructive sampling every 7–15 days, the AGB samples from which were used as the ground truth data. We evaluated the ability of the ML models to estimate the real-time values of AGB at a sub-meter resolution (0.17–0.52 m2). An extremely randomized trees (ERT) regressor was selected for the regression analysis, based on its predictive performance for the first year’s growing season. The model was retrained using previously identified hyperparameters to predict the AGB of the crops in the second year. The ERT performed AGB estimation using height and reflectance metrics from LiDAR-derived point cloud data and achieved a prediction performance of R2 = 0.48 at a spatial resolution of 0.35 m2. The prediction performance could be improved significantly by aggregating adjacent predictions (R2 = 0.71 and R2 = 0.93 at spatial resolutions of 1 m2 and 2 m2, respectively) as they ultimately converged to the reference biomass values because any individual errors averaged out. The AGB prediction results were examined as function of predictor type, training set size, sampling resolution, phenology, and canopy density. The results demonstrated that when combined with ML regression methods, the UAV–LiDAR method could be used to provide accurate real-time AGB prediction for crop fields at a high resolution, thereby providing a way to map their biochemical constituents.

Improved AdaBoost algorithm using misclassified samples oriented feature selection and weighted non-negative matrix factorization

To improve the classification performance of existing adaptive boosting (AdaBoost) based algorithms effectively, an improved AdaBoost algorithm based on misclassified samples oriented feature selection and weighted non-negative matrix factorization (WNMF) is proposed in this paper. Firstly, in order to consider the effects of sample weights, a misclassified samples oriented feature selection (called MOFS) is proposed to select the most discriminative features which occur in the samples with high weights. Secondly, the explicit features and the part-based features of the training samples are both considered, and the WNMF algorithm is introduced and combined with MOFS to reduce the dimension of the training sample set. Finally, the concept of misclassification degree is introduced and a fine grained sample weight updating method is proposed to distinguish the samples with different misclassification degrees. Numerical experiments show that the proposed MOFS method achieves higher accuracy compared to traditional feature selection methods, and the proposed MOFS and WNMF based AdaBoost method obtains significant improvement on classification accuracy when comparing with typical existing AdaBoost based algorithms using different classifiers.

Diagnosis method based on hidden Markov model and Weibull mixture model for mechanical faults of in-wheel motors

To effectively monitor the operation state of in-wheel motors used in electric vehicles and ensure the safety of the whole vehicle, a diagnosis method based on hidden Markov model (HMM) and Weibull mixture model (WMM) is proposed for mechanical faults in in-wheel motors, known simply as the WMM-HMM diagnosis method. Firstly, vibration signals of the in-wheel motor are extracted for sensitive symptom parameters which are used to characterize the operation state and establish the observation sequence. Secondly, WMM is employed to expand the limited observation sequence under various operating states of in-wheel motors to obtain sufficient observation sequence as the training sample set of HMM, and HMM parameters are determined through combining supervised learning with unsupervised learning algorithm. Then the WMM-HMM diagnosis models are constructed under low and medium speed conditions respectively. Finally, the corresponding faults in-wheel motors are customized and the test bench is built to verify the proposed method. The test results show that the proposed method can accurately identify the mechanical fault state of in-wheel motors under different conditions and has good generalization and applicability in traditional methods comparison.

Comparison of the predictive performance of Logistic regression, BP neural network and support vector machine model for the risk of acute exacerbation of readmission in elderly patients with chronic obstructive pulmonary disease within 30 days

To compare the effectiveness of Logistic regression, BP neural network and support vector machine models in the prediction of 30-day risk of readmission in elderly patients with an exacerbation of chronic obstructive pulmonary disease (COPD) and to provide a scientific basis for the screening and prevention of high-risk patients with readmission. The COPD patient survey questionnaire was made, including the general data questionnaire, modified Medical Research Council dyspnea scale (mMRC), activities of daily living (ADL), the geriatric depression scale, the mini nutritional assessment-short form (MNA-SF) and COPD assessment test (CAT). Elderly COPD patients were selected from the department of respiratory medicine of 13 general hospitals in Ningxia from April 2019 to August 2020 by convenience sampling method, and they were followed up 30 days after discharge. To explore the risk factors of patient readmission, Logistic regression model, BP neural network model and support vector machine models were constructed based on the risk factors. According to the ratio of the training set to the testing set of 7:3, the model was divided into the training set sample and the testing set sample. The prediction efficiency of the model was compared by the precision rate, recall rate and accuracy rate, F1 index and the area under the receiver operator characteristic curve (AUC). A total of 1 120 patients were investigated, including 879 non-readmission patients and 241 readmission patients. Univariate regression analysis showed that there were statistically significant differences in age, education level, smoking status, proportion of diabetes and coronary heart disease, hospitalization times of acute exacerbation of COPD in the past 1 year, seasonal factors and long-term home oxygen therapy, regular medication, proportion of rehabilitation exercise, course of disease, ADL, depression status, mMRC, nutritional status between non-readmission patients and readmission patients. Binary Logistic regression analysis showed that education level, smoking status, coronary heart disease, hospitalization times of acute exacerbation of COPD in the past 1 year, seasonal factors, whether long-term home oxygen therapy, whether regular medication, nutritional status were the risk factors for 30-day acute exacerbation of readmission in elderly patients with COPD. The training set showed that the accuracy rate of Logistic regression model, BP neural network model and support vector machine models were 70.95%, 76.51% and 84.78%, respectively. The recall rates were 79.55%, 86.36% and 88.64%, respectively. The accuracy rates were 87.81%, 90.81% and 93.82%, respectively. F1 indexes were 0.75, 0.81 and 0.87, respectively. The AUC were 0.850, 0.893 and 0.921, respectively. The testing set showed that the precision rate of Logistic regression model, BP neural network model and support vector machine model were 78.38%, 80.65% and 88.57%, respectively. The recall rates were 70.73%, 60.98% and 75.61%, respectively. The accuracy rates were 85.82%, 84.40% and 90.07%, respectively. F1 indexes were 0.74, 0.69 and 0.82, respectively. The AUC were 0.814, 0.775 and 0.858, respectively. Comparing with Logistic regression and BP neural network, support vector machine model has better prediction effect, and can effectively predict the risk of acute exacerbation of readmission in elderly patients with COPD within 30 days.

A fingerprint-based localization algorithm based on LSTM and data expansion method for sparse samples

The accuracy of WiFi fingerprint-based localization is related to the number of reference points, generally, to obtain better positioning accuracy, enough samples must be collected, which will inevitably lead to a huge sampling workload. Thus, it will be of great significance to design an algorithm using sparse samples to achieve positioning accuracy like that of dense samples. This paper proposes a WiFi fingerprint-based localization algorithm using Long Short-Term Memory Network (LSTM) with explainable feature and a sparse sample expansion algorithm (PGSE) based on Principal component analysis and Gaussian process regression for sparse samples. Specifically, in the case of limited number of collected reference points, principal component analysis is used to select the access point, and Gaussian process regression is used to model the reference point coordinates and the corresponding received signal strength values in the training sample set, to expand the signal data and construct a new fingerprint database. The effectiveness of the PGSE algorithm is verified by using the public dataset ’UJIIndoorLoc’. At the same time, the applicability of PGSE expansion algorithm to data with temporal information is verified in the fingerprint-based localization method. In addition, this paper also proposes a WiFi-RSSI indoor localization method based on Long Short-Term Memory Network. Lots of experiments are conducted in the actual scenes and the results are compared with several existing methods. The results indicate that the proposed method improves the precision of indoor localization on an average level compared to state-of-the-art methods.