Optimal Classification Research Articles

Optimized Deep CNN with Deviation Relevance-based LBP for Skin Cancer Detection: Hybrid Metaheuristic Enabled Feature Selection

Segmentation of skin lesions is a significant and demanding task in dermoscopy images. This paper proposes a new skin cancer recognition scheme, with: “Pre-processing, Segmentation, Feature extraction, Optimal Feature Selection and Classification”. Here, pre-processing is done with certain processes. The pre-processed images are segmented via the “Otsu Thresholding model”. The third phase is feature extraction, where Deviation Relevance-based “Local Binary Pattern (DRLBP), Gray-Level Co-Occurrence Matrix (GLCM) features and Gray Level Run-Length Matrix (GLRM) features” are extracted. From these extracted features, the optimal features are chosen via Particle Updated WOA (PU-WOA) model. Subsequently, classification occurs via Optimized DCNN and NN to classify the skin lesion. To make the classification more precise, the DCNN is optimized by the introduced algorithm. The result has shown a higher accuracy of 0.998737, when compared with other extant models like IPSO, IWOA, PSO+CNN, WOA+CNN and CNN schemes.

Fuzzy Neighbors and Deep Learning-Assisted Spark Model for Imbalanced Classification of Big Data

Big data is important in knowledge manipulation, assessment, and prediction. However, extracting and analyzing knowledge through big database are complex because of imbalance data distribution that leads to wrong decisions and biased classification outputs. Hence, an effective and optimal big data classification approach is designed using the proposed Bird Swarm Deer Hunting Optimization-Deep Belief Network (BSDHO-based DBN) algorithm based on spark architecture that follows the master and slave nodes. The proposed BSDHO is obtained by combining Deer Hunting Optimization algorithm and Bird Swarm Algorithm. The developed model poses two nodes, namely slave and master node. The training data is initially given to the master node in the spark architecture to perform transformation of data. Here, the transformation of data is done with an exponential log kernel, and then selection of feature is done with sequential forward selecting for choosing suitable features for enhanced processing. Consequently, oversampling process is performed with Fuzzy K-Nearest Neighbor (Fuzzy KNN) in the slave node using selected features to manage imbalance data. Then, in master node, classification is done with Deep belief Network, and trained using developed Bird swarm Deer Hunting Optimization (BSDHO) algorithm. On the other hand, the test data is taken as input, and is fed to the slave node to perform data transformation. Then, the transformed data is given to the master node for classification based on the proposed BSDHO. At last, the training data and testing data output produced the classified output. The proposed BSDHO-based DBN provided enhanced outcomes with highest specificity of 97.92%, accuracy of 96.92%, and sensitivity of 96.9%.

Hybrid Classifier-Based Federated Learning in Health Service Providers for Cardiovascular Disease Prediction

One of the deadliest diseases, heart disease, claims millions of lives every year worldwide. The biomedical data collected by health service providers (HSPs) contain private information about the patient and are subject to general privacy concerns, and the sharing of the data is restricted under global privacy laws. Furthermore, the sharing and collection of biomedical data have a significant network communication cost and lead to delayed heart disease prediction. To address the training latency, communication cost, and single point of failure, we propose a hybrid framework at the client end of HSP consisting of modified artificial bee colony optimization with support vector machine (MABC-SVM) for optimal feature selection and classification of heart disease. For the HSP server, we proposed federated matched averaging to overcome privacy issues in this paper. We tested and evaluated our proposed technique and compared it with the standard federated learning techniques on the combined cardiovascular disease dataset. Our experimental results show that the proposed hybrid technique improves the prediction accuracy by 1.5%, achieves 1.6% lesser classification error, and utilizes 17.7% lesser rounds to reach the maximum accuracy.

Radiomics-based intracranial thrombus features on preoperative noncontrast CT predicts successful recanalization of mechanical thrombectomy in acute ischemic stroke.

To evaluate the predictive value of radiomics features extracted from the thrombus on preoperative computed tomography images to identify successful recanalization after stent retrieve (SR) treatment in patients with acute ischemic stroke (AIS). Two hundred fifty-six patients newly diagnosed AIS between March 2017 and September 2020 from two institutes, including the first affiliated hospital of Soochow university (institute I) and Northern Jiangsu People's hospital (institute II), were enrolled continuously and retrospectively. Patients with unsatisfactory image quality were excluded. The remaining patients of institute I were randomly divided into the training and internal validation cohorts at a ratio of 7 to 3, and patients of institute II were collected as the external validation cohort. After extraction and selection of the optimal radiomics features from training cohort, six machine learning (ML) classifiers including naïve Bayes (NB), random forest (RF), logistic regression (LR), linear support vector machine (L.SVM), radial SVM (R.SVM), and an artificial neural network (ANN) were developed to predict successful recanalization with SR treatment and compared. A combined model based on the optimal ML classifier was constructed using the optimal radiomics model and clinical-radiological risk variables. Finally, the performance of the model was selected based on the Matthews correlation coefficient (MCC) and the area under the receiver operating (AUC) and independently evaluated on the internal validation and external validation cohorts. We automatically extracted 1,130 radiomics features from the voxel of interest (VOI) using PyRadiomics. The eight most relevant radiomics features were identified using Intraclass coefficient, single-factor logistic regression analysis, and least absolute shrinkage and selection operator algorithm in the training cohort. Among the six ML classifiers, the ANN classifier using thrombus radiomics features achieved the best prediction of early recanalization under SR with MCCs of 0.913, 0.693 and 0.505 in training, internal and external validation cohorts, respectively. Moreover, receiver operating characteristic curves showed that the combined model [AUC =0.860, 95% confidence interval (CI): 0.731-0.936; AUC =0.849, 95% CI: 0.759-0.831] was not significantly better than radiomics model based on the ANN classifier alone (AUC =0.873, 95% CI: 0.803-0.891; AUC =0.805, 95% CI: 0.864-0.971) (P>0.05, Delong test) in internal and external validation cohorts. A radiomics model based on the ANN classifier has the ability to predict successful recanalization after SR in patients with AIS, thus allowing a potentially better selection of mechanical thrombectomy treatment.

Optimal Ensemble Scheme for Human Activity Recognition and Floor Detection Based on AutoML and Weighted Soft Voting Using Smartphone Sensors

In this article, human activity recognition (HAR) attempts to recognize the activities of an object in a multistory building from data retrieved via smartphone-based sensors (SBSs). Most publications based on machine learning (ML) report the development of a suitable architecture to improve classification accuracy by increasing the parameters of the architecture pertaining to HAR. Due to robust and automated ML, it is quite possible to develop a versatile approach to improve the accuracy of HAR. This research proposes an optimal ensemble HAR and floor detection (OEC-HAFD) scheme based on automated learning (AutoML) and weighted soft voting (WSF) using SBS to improve the recognition rate. The proposed HAR-FD scheme is developed based on twofold mechanisms. First, an AutoML paradigm is employed to find optimal supervised models based on performance. Second, top-ranked (optimal) models are combined using the WSF mechanism to classify HAR on various floors of a multistory building. The proposed scheme is developed based on real-time SBS: accelerometer and barometer data. The accelerometer data are used to detect activity by observing the magnitude of the sensor measurements. Similarly, the barometer sensor detects floor height by using pressure and altitude data. Furthermore, we analyze the performance of the proposed optimal ensemble classifier (OEC) and compare them with the state-of-the-art classifiers. Based on the result analysis, it is clearly seen that the proposed OEC outperforms the performance of the individual classifiers. Moreover, our proposed HAR-FD can be leveraged as a robust solution to accurately recognize human activities in multistory buildings compared to the existing standalone models.

Machine Learning Quantified Tumor-Stroma Ratio Is an Independent Prognosticator in Muscle-Invasive Bladder Cancer.

Although the tumor-stroma ratio (TSR) has prognostic value in many cancers, the traditional semi-quantitative visual assessment method has inter-observer variability, making it impossible for clinical practice. We aimed to develop a machine learning (ML) algorithm for accurately quantifying TSR in hematoxylin-and-eosin (H&E)-stained whole slide images (WSI) and further investigate its prognostic effect in patients with muscle-invasive bladder cancer (MIBC). We used an optimal cell classifier previously built based on QuPath open-source software and ML algorithm for quantitative calculation of TSR. We retrospectively analyzed data from two independent cohorts to verify the prognostic significance of ML-based TSR in MIBC patients. WSIs from 133 MIBC patients were used as the discovery set to identify the optimal association of TSR with patient survival outcomes. Furthermore, we performed validation in an independent external cohort consisting of 261 MIBC patients. We demonstrated a significant prognostic association of ML-based TSR with survival outcomes in MIBC patients (p < 0.001 for all comparisons), with higher TSR associated with better prognosis. Uni- and multivariate Cox regression analyses showed that TSR was independently associated with overall survival (p < 0.001 for all analyses) after adjusting for clinicopathological factors including age, gender, and pathologic stage. TSR was found to be a strong prognostic factor that was not redundant with the existing staging system in different subgroup analyses (p < 0.05 for all analyses). Finally, the expression of six genes (DACH1, DEEND2A, NOTCH4, DTWD1, TAF6L, and MARCHF5) were significantly associated with TSR, revealing possible potential biological relevance. In conclusion, we developed an ML algorithm based on WSIs of MIBC patients to accurately quantify TSR and demonstrated its prognostic validity for MIBC patients in two independent cohorts. This objective quantitative method allows application in clinical practice while reducing the workload of pathologists. Thus, it might be of significant aid in promoting precise pathology services in MIBC.

Nearest Neighbor-Based Strategy to Optimize Multi-View Triplet Network for Classification of Small-Sample Medical Imaging Data.

Multi-view classification with limited sample size and data augmentation is a very common machine learning (ML) problem in medicine. With limited data, a triplet network approach for two-stage representation learning has been proposed. However, effective training and verifying the features from the representation network for their suitability in subsequent classifiers are still unsolved problems. Although typical distance-based metrics for the training capture the overall class separability of the features, the performance according to these metrics does not always lead to an optimal classification. Consequently, an exhaustive tuning with all feature-classifier combinations is required to search for the best end result. To overcome this challenge, we developed a novel nearest-neighbor (NN) validation strategy based on the triplet metric. This strategy is supported by a theoretical foundation to provide the best selection of the features with a lower bound of the highest end performance. The proposed strategy is a transparent approach to identify whether to improve the features or the classifier. This avoids the need for repeated tuning. Our evaluations on real-world medical imaging tasks (i.e., radiation therapy delivery error prediction and sarcoma survival prediction) show that our strategy is superior to other common deep representation learning baselines [i.e., autoencoder (AE) and softmax]. The strategy addresses the issue of feature's interpretability which enables more holistic feature creation such that the medical experts can focus on specifying relevant data as opposed to tedious feature engineering.

Liver Disease Detection

In recent times, intelligent predictive systems are showing greater levels of accuracy and effectiveness in early detection of the critical diseases of cancer in the liver, lungs, etc. Predictive models assist medical practitioners to identify the diseases based on symptoms and health indicators like hormones, enzymes, age, blood counts, etc. This article focuses on proposing an optimal classification model to detect chronic liver disease by enhancing the prediction accuracy through cutting-edge analytics. The article proposes an enhanced framework on the original study by Ramana et al. It uses measures like precision and balanced accuracy to choose the most efficient classification algorithm in Indian and USA patient datasets using various factors like enzymes, age, etc. Using Youden's index, individual thresholds for each model were identified to increase the power of sensitivity and specificity, respectively. The study proposes a framework for highly accurate automated disease detection in the medical industry and helps in strategizing preventive measures for patients.

Potential plasma biomarker panels identification for the diagnosis of first-episode schizophrenia and monitoring antipsychotic monotherapy with the use of metabolomics analyses

Schizophrenia (SCZ) is a severe mental disorder. Using liquid chromatography mass spectrometry, we performed comprehensive metabolomics analyses of plasma samples from healthy controls (HC) and first episode SCZ patients before and after an acute period of medication. Ten lipid metabolites and 27 soluble small molecules were identified as potential biomarkers associated with the diagnosis and treatment of SCZ. These metabolites were significantly reduced in SCZ, and lipids and sulfate were significantly increased after treatment. Of the metabolites identified, four showed significant correlations with the Positive and Negative Syndrome Scale total scores. A biomarker panel composed of alpha-dimorphecolic, Phosphatidylcholine (PC) (16:0/18:1(11Z)), 1-methylnicotinamide, Phosphatidylethanolamine (PE) (20:2(11Z,14Z)/18:2(9Z,12Z)), sulfate, and L-tryptophan was selected to distinguish SCZ from HC; this provided the maximum classification performance with an AUC of 0.972. A biomarker panel including C16 sphinganine, gamma-linolenic acid, linoleic acid, PC(16:0/18:1(11Z)), PE(20:2(11Z,14Z)/18:2(9Z,12Z)), and sulfate, was selected for discrimination between SCZ before and after medication, and produced the optimal classification performance with an AUC of 0.905. Disturbances in lipid metabolism, sulfation modification, tryptophan metabolism, anti-inflammatory and antioxidant systems, and unsaturated fatty acids metabolism, were identified in SCZ. Our findings could facilitate the development of objective diagnostic or drug treatment monitoring tools for schizophrenia.

Efficient learning of large sets of locally optimal classification rules

Conventional rule learning algorithms aim at finding a set of simple rules, where each rule covers as many examples as possible. In this paper, we argue that the rules found in this way may not be the optimal explanations for each of the examples they cover. Instead, we propose an efficient algorithm that aims at finding the best rule covering each training example in a greedy optimization consisting of one specialization and one generalization loop. These locally optimal rules are collected and then filtered for a final rule set, which is much larger than the sets learned by conventional rule learning algorithms. A new example is classified by selecting the best among the rules that cover this example. In our experiments on small to very large datasets, the approach’s average classification accuracy is higher than that of state-of-the-art rule learning algorithms. Moreover, the algorithm is highly efficient and can inherently be processed in parallel without affecting the learned rule set and so the classification accuracy. We thus believe that it closes an important gap for large-scale classification rule induction.

Singular solutions for the fourth‐order parabolic equation with nonstandard growth conditions and absorption

This paper deals with the fourth‐order parabolic equation in a bounded domain, subject to homogeneous Navier boundary conditions. Under some conditions on the variable exponents, we give a complete and optimal classification on the singularity of solutions, characterized by the signs of the Nehari energy and the difference between the initial energy and the potential depth.

Optimized RNN-based performance prediction of IoT and WSN-oriented smart city application using improved honey badger algorithm

Internet of Things (IoT) has been widely utilized as the major significant element of Information and Communications Technology (ICT) for feasible smart cities due to the capability of IoT for supporting sustainability in numerous domains. There is a need for fault avoidance via the continuous and dynamic utilization of network behavior for achieving the necessary quality of the IoT communication system that allows us to get sustainable enhancement in smart cities regarding IoT communication systems. While considering the IoT-assisted network, the basic part of the IoT model is considered a Wireless Sensor Network (WSN), especially for data management that consists of various sensor nodes in the smart city area. Moreover, every node in a WSN is considerably utilized for a specific reason and thus, every node is handled by a battery that increases the consumption of energy in the entire smart city scheme for processing the data regarding communication. On the other hand, there exist various limitations like scalability, communication latency, centralization, privacy, security, etc. Thus, this paper plans to develop the IoT and WSN-based smart city application using novel intelligent techniques, aiming for the optimal performance classification of the network. The infrastructure of IoT will consist of WSN, Vehicular Ad-hoc NETwork (VANET), Mobile Ad-hoc NETworks (MANET), Radio Frequency Identification (RFID), and Wireless Body Area Networks (WBAN). Here, the efficiency of the whole IoT system is acquired from the efficiency of the whole IoT module. Hence, the IoT network efficiency rates of the individual must be predicted in an earlier stage to predict the effectiveness of the complete IoT system. The efficiency of each network is forecasted with the help of the constraints such as energy consumption, gathered data size, transmitted data size, mobility, false positive, throughput, packet loss, and delay. The output from each network is considered as input to the Optimized Recurrent Neural Network (ORNN) for predicting the ending superiority of the whole IoT network. Here, the parameter tuning in the RNN is done by the Self Adaptive Honey Badger Algorithm (SA-HBA). The empirical outcomes confirmed that the designed method has forecasted and enhanced the superiority of a whole simulated IoT system in an accurate manner. Throughout the result analysis, the given designed method attains minimum energy consumption rate and also better prediction accuracy.

Application of the AT(N) and Other CSF Classification Systems in Behavioral Variant Frontotemporal Dementia.

Patients with a frontotemporal lobar degeneration (FTLD) usually manifest with behavioral variant frontotemporal dementia (bvFTD). Alzheimer's disease (AD) may also manifest with a predominant behavioral-dysexecutive syndrome, similar to bvFTD. Cerebrospinal fluid (CSF) biomarkers, such as total tau (τT), phosphorylated tau (τP-181) and amyloid beta with 42 amino-acids (Aβ42), can predict AD pathology in vivo. The aim of this study was to compare the τT/Aβ42 and τP-181/Aβ42 ratios, the BIOMARKAPD/ABSI criteria and the AT(N) classification system in a cohort of bvFTD patients. A total of 105 bvFTD patients (21 possible bvFTD; 20%) with CSF data, examined from 2008 to 2022, were included. Seventy-eight AD patients and 62 control subjects were included. The CSF biomarkers were measured with Innotest (2008-2017 subcohort) and EUROIMMUN (2017-2022 subcohort) ELISAs. Depending on the classification system, 7.6 to 28.6% of bvFTD had an AD biochemical profile. The τT/Aβ42 and τP-181/Aβ42 ratios classified more patients as AD compared to the BIOMARKAPD/ABSI and AT(N) systems. The patients with possible bvFTD had higher frequencies of AD compared to the probable bvFTD patients. The four classification criteria of CSF AD biomarkers resulted in differences in AD allocation in this bvFTD cohort. A consensus on the optimal classification criteria of CSF AD biomarkers is pivotal.

Impact of high-risk features on outcome of acute type B aortic dissection

BackgroundAcute type B aortic dissection (TBAD) is a severe condition associated with significant morbidity and mortality. The optimal classification and treatment strategy of TBAD remain controversial and inconsistent. MethodsThis analysis includes patients treated for acute TBAD at the Helsinki University Hospital, Finland between 2007 and 2019. The endpoints were early and late mortality, intervention of the aorta, and a composite of death and aortic intervention in uncomplicated patients and high-risk patients. ResultsThis study included 162 consecutive TBAD patients (27.8% females), 114 in the high-risk group and 48 in the uncomplicated group, with a mean age of 67.6 ± 13.9 years. Intramural hematoma was reported in 63 cases (38.9%). The mean follow-up was 5.1 ± 3.9 years. In-hospital/30-day mortality (n = 4; 3.5%) occurred solely in the high-risk group (P = .32). Additionally, TBAD-related adverse events (n = 23; 20.2%) were observed only in the high-risk group (P < .001). The cumulative incidences of the composite TBAD outcome with non–TBAD-related death as a competing risk were 6.6% (95% CI, 1.7%-16.5%) in the uncomplicated group and 29.5% (95% CI, 21.1%-38.3%) in the high-risk group at 5 years and 6.6% (95% CI, 1.7%-16.5%) and 33.0% (95% CI, 23.7%-42.6%) at 10 years (P = .001, Gray test). Extracardiac arteriopathy (subdistribution hazard ratio [SHR], 2.61; 95% CI, 1.08-6.27) and coronary artery disease (SHR, 2.24; 95% CI, 1.07-4.71) were risk factors for adverse aortic-related events in univariable competing-risk regression analysis. ConclusionsRecognition of risk factors underlying adverse events related to TBAD is essential because the disease progression impacts both early and late outcomes. Early aortic repair in high-risk TBAD may reduce long-term morbidity and mortality.

Evaluating the predictive performance of subtyping: Acriterion for cluster mean-based prediction.

Heterogeneity is a frequent issue in population data analyses in medicine, biology, and the social sciences. A common approach for handling heterogeneity is to use a clustering algorithm to group similar samples, considering samples within the same group to be homogeneous. This approach is known as "subtyping" or "subgrouping." Methods for evaluating the validity of subtyping have yet to be fully established. In this study, we propose the cost of cluster mean-based prediction (CCMP) as a metric for evaluating the accuracy of predictions based on subtyping. By selecting the minimum CCMP among several candidate clustering results, the optimal subtype classification in terms of prediction accuracy can be determined. The computational implementation of the CCMP is validated with numerical experiments. We also examine some properties of subtype classification selected by CCMP.

Automatic identification of conodont species using fine-grained convolutional neural networks

Conodonts are jawless vertebrates deposited in marine strata from the Cambrian to the Triassic that play an important role in geoscience research. The accurate identification of conodonts requires experienced professional researchers. The process is time-consuming and laborious and can be subjective and affected by the professional level and opinions of the appraisers. The problem is exacerbated by the limited number of experts who are qualified to identify conodonts. Therefore, a rapid and simple artificial intelligence method is needed to assist with the identification of conodont species. Although the use of deep convolutional neural networks (CNN) for fossil identification has been widely studied, the data used are usually from different families, genera or even higher-level taxonomic units. However, in practical geoscience research, geologists are often more interested in classifying species belonging to the same genus. In this study, we use five fine-grained CNN models on a dataset consisting of nine species of the conodont genus Hindeodus. Based on the cross-validation results, we show that using the Bilinear-ResNet18 model and transfer learning generates the optimal classifier. Area Under Curve (AUC) value of 0.9 on the test dataset was obtained by the optimal classifier, indicating that the performance of our classifier is satisfactory. In addition, although our study is based on a very limited taxa of conodonts, our research principles and processes can be used as a reference for the automatic identification of other fossils.

Tree Species Classification Using Airborne LiDAR Data Based on Individual Tree Segmentation and Shape Fitting

Individual tree species classification is of strategic importance for forest monitoring, analysis, and management, which are critical for sustainable forestry development. In this regard, the paper proposes a method based on the profile of segmented individual tree laser scanning points to identify tree species. The proposed methodology mainly takes advantage of three-dimensional geometric features of a tree crown captured by a laser point cloud to identify tree species. Firstly, the Digital Terrain Model (DTM) and Digital Surface Model (DSM) are used for Crown Height Model (CHM) generation. Then, local maximum algorithms and improved rotating profile-based delineations are used to segment individual trees from the profile CHM point data. In the next step, parallel-line shape fitting is used to fit the tree crown shape. In particular, three basic geometric shapes, namely, triangle, rectangle, and arc are used to fit the tree crown shapes of different tree species. If the crown belongs to the same crown shape or shape combination, parameter classification is used, such as the ratio of crown width and crown height or the apex angle range of the triangles. The proposed method was tested by two real datasets which were acquired from two different sites located at Tiger and Leopard National Park in Northeast China. The experimental results indicate that the average tree classification accuracy is 90.9% and the optimal classification accuracy reached 95.9%, which meets the accuracy requirements for rapid forestry surveying.

Application of portable near‐infrared spectroscopy technology for grade identification of Panax notoginseng slices

AbstractPanax notoginseng slices (PNS) are prepared from the taproot of a rare Chinese herbal plant, Panax notoginseng. The price and efficacy of PNS change depending on its grade, but substandard PNS are more prevalent in the Asian market. In this study, a portable near‐infrared spectrometer was used to collect the spectra of 240 PNS samples divided into four grades. The spectral data were preprocessed by the Savitzky–Golay (SG) filter to eliminate noise interference. Principal component analysis (PCA), competitive adaptive reweighted sampling, and variable combination population analysis were used to extract the feature variables of the spectral data. The selected feature variables were used to establish least squares support vector machine (LSSVM), support vector machine (SVM), and extreme learning machine (ELM) classification models. To further improve the classification accuracy of the most effective of these models, a grey wolf optimizer (GWO) was introduced, and particle swarm optimization (PSO) as well as genetic algorithm (GA) were used to conduct comparative analyses. The results showed that PCA provided accurate identification information of different PNS grades and that the classification effect of the LSSVM model was better than that of the ELM and SVM models. During the optimization process, the optimization accuracy of the GWO was better than that of the PSO and GA systems. Therefore, the optimal classification model was established as GWO–PCA–LSSVM, and the classification accuracy of the test set was 91.67%. Therefore, portable near‐infrared spectroscopy technology can be used to identify the grade of PNS effectively.

Development of surrogate models for evaluating energy transfer quality of high-speed railway pantograph-catenary system using physics-based model and machine learning

High-speed railway pantograph-catenary system is the only energy transfer pathway to drive a train operation. Energy transfer quality deteriorates with the increasing train speed and harsh service environment, thereby quickly and accurately evaluating the energy transfer quality is very important to guarantee the normal operation of a train. In this study, firstly, the physics-based model to simulate the dynamic interaction of pantograph-catenary system is established and validated. Eleven input parameters involve the essential line design and train operation parameters, and the output parameters that are crucially responsible for energy transfer quality are obtained by feature extraction. Secondly, a sampling strategy is employed to construct the input sampling points, based on which the outputs are computed via physics-based model, then combining them the dataset is obtained. Thirdly, five tree-based classification surrogate models are developed and compared to assess the level of energy transfer quality. Finally, eight regression surrogate models are developed in replacing physics-based model to evaluate the essential values of energy transfer quality. It is found that the gradient boosting decision tree (GBDT)-based surrogate model is the optimal classification model and the multi-layer feed-forward deep neural network (MLF-DNN)-based surrogate model for the optimal regression model. The two surrogate models are expected to quickly find the optimal design parameters and improve the operation control of trains of high-speed railway for the purpose of enhancing the energy transfer quality if coupled with optimization procedure.

A novel hybrid gene selection for tumor identification by combining multifilter integration and a recursive flower pollination search algorithm

Gene selection is crucial to tumor identification based on microarray expression data. The identification of genes with strong discriminative power has been a hot research topic and a great challenge in research related to machine learning-based tumor identification. This study proposes a novel hybrid gene selection strategy for tumor identification by combining multifilter integration and a recursive flower pollination algorithm (MFI-RFPA) to select a minimum number of genes to achieve optimal classification performance. MFI-RFPA selects genes that involve two stages. In the first stage, most irrelevant and noise genes are eliminated by using a combination of three filters based on different metrics. In this process, multiple filters based on different metrics are combined in the voting, which makes the filter process of MFI-RFPA more robust. In the second stage, MFI-RFPA searches for the optimal subset of the remaining genes by using a recursive flower pollination algorithm (RFPA). In this process, the flower pollination algorithm is performed recursively without reducing the classification accuracy, which ensures that a small and compact subset of genes with excellent classification performance is searched. In the experimental part, MFI-RFPA is compared with both its directly related methods (such as MFI, FPA, RFPA, MFI-FPA) and other state-of-the-art methods. The excellent performance of MFI-RFPA in reducing the number of genes and improving the classification accuracy can be reflected in the comparison of experiments.