Feature Reduction Algorithm Research Articles

Computational approach for decoding malaria drug targets from single-cell transcriptomics and finding potential drug molecule

Malaria is a deadly disease caused by Plasmodium parasites. While potent drugs are available in the market for malaria treatment, over the years, Plasmodium parasites have successfully developed resistance against many, if not all, front-line drugs. This poses a serious threat to global malaria eradication efforts, and the continued discovery of new drugs is necessary to tackle this debilitating disease. With recent unprecedented progress in machine learning techniques, single-cell transcriptomic in Plasmodium offers a powerful tool for identifying crucial proteins as a drug target and subsequent computational prediction of potential drugs. In this study, We have implemented a mutual-information-based feature reduction algorithm with a classification algorithm to select important proteins from transcriptomic datasets (sexual and asexual stages) for Plasmodium falciparum and then constructed the protein-protein interaction (PPI) networks of the proteins. The analysis of this PPI network revealed key proteins vital for the survival of Plasmodium falciparum. Based on the function and identification of a few strong binding sites on a couple of these key proteins, we computationally predicted a set of potential drug molecules using a deep learning-based technique. Lead drug molecules that satisfy ADMET and drug-likeliness properties are finally reported out of the generated drugs. The study offers a general computational pipeline to identify crucial proteins using scRNA-seq data sets and further development of potential new drugs.

An enhanced algorithm for semantic-based feature reduction in spam filtering.

With the advent and improvement of ontological dictionaries (WordNet, Babelnet), the use of synsets-based text representations is gaining popularity in classification tasks. More recently, ontological dictionaries were used for reducing dimensionality in this kind of representation (e.g., Semantic Dimensionality Reduction System (SDRS) (Vélez de Mendizabal et al., 2020)). These approaches are based on the combination of semantically related columns by taking advantage of semantic information extracted from ontological dictionaries. Their main advantage is that they not only eliminate features but can also combine them, minimizing (low-loss) or avoiding (lossless) the loss of information. The most recent (and accurate) techniques included in this group are based on using evolutionary algorithms to find how many features can be grouped to reduce false positive (FP) and false negative (FN) errors obtained. The main limitation of these evolutionary-based schemes is the computational requirements derived from the use of optimization algorithms. The contribution of this study is a new lossless feature reduction scheme exploiting information from ontological dictionaries, which achieves slightly better accuracy (specially in FP errors) than optimization-based approaches but using far fewer computational resources. Instead of using computationally expensive evolutionary algorithms, our proposal determines whether two columns (synsets) can be combined by observing whether the instances included in a dataset (e.g., training dataset) containing these synsets are mostly of the same class. The study includes experiments using three datasets and a detailed comparison with two previous optimization-based approaches.

Detection of Geographic Information System Security Hazards in the IoT Based on Network Security Situation Awareness

Abstract The geographic information system (GIS) is the important system these days for navigation and directing the routes. In order to detect the security risks of the GIS efficiently and accurately and to protect the security of the GIS, a GIS security risk detection method in the Internet of Things based on network security situational awareness has been proposed. The data preprocessing method based on the operation data of the GIS is adopted. The normalized and standardized processing of the operation data of the GIS is devised to reduce the dimensional difference between the data and the geographical areas. This will facilitate the subsequent applications of the GIS to make better decisions by preserving the security. The redundant data in the GIS operation data are removed by using feature reduction algorithms through the acquisition method. The security elements based on enhanced probabilistic neural network are inferred, and the useful data samples are obtained as the security elements detection set. A rough set theory is also used to realize geographic security hidden danger identification. The experimental results show that the proposed method is efficient to maintain security and to detect the geographic security risk detection within a short span of time. The detection result accuracy is high, which meets the GIS application requirements.

Investigating the feasibility of differentiating MS active lesions from inactive ones using texture analysis and machine learning methods in DWI images

BackgroundMagnetic resonance imaging (MRI) is commonly used in conjunction with a gadolinium-based contrast agent (GBCA) to distinguish active multiple sclerosis (MS) lesions. However, recent studies have raised concerns regarding the long-term effects of the accumulation of GBCA in the body. Thus, the purpose of this study is to investigate the possibility of using texture analysis in diffusion-weighted imaging (DWI) and machine learning algorithms to discriminate active from inactive MS lesions without the use of GBCA. MethodsTo achieve this purpose, we introduce an image processing pipeline. In the proposed pipeline, following registration and alignment of slices, MS lesions from DWI images are segmented and quantized. Next, different texture analysis methods are employed to extract texture features from the lesions. Then, a two-stage feature reduction method is applied, in which the first stage involves a statistical t-test and the second stage relies on principal component analysis (PCA), sequential forward selection (SFS), sequential backward selection (SBS), and ReliefF algorithms. Finally, we use five classifiers logistic regression (LR), support vector machine (SVM), decision tree (DT), K nearest neighbor (KNN), and linear discriminant analysis (LDA) in a 5-fold cross-validation procedure to determine active and inactive MS lesions. ResultsIn this study, we collected and prepared 255 active/inactive MS lesions from MRI scans of 34 patients diagnosed with MS, with a mean age of 35.56±10.89. Among 89 texture features extracted, 63 features showed statistically significant differences between the means of active and inactive lesions (P<0.05). The SVM classifier with the PCA feature reduction algorithm demonstrated the best performance with an average accuracy of 0.960 (±0.024), specificity and precision of 1.0, sensitivity of 0.913 (±0.053), and AUC of 0.957 (±0.027). ConclusionOur study indicates that DWI changes detected using texture analysis-based machine learning models can precisely differentiate active from inactive MS lesions. This finding provides valuable clinical information for the early diagnosis and effective monitoring of MS disease.

A Network Intrusion Detection Method Based on a Fast KNN Algorithm

This paper proposes a network intrusion detection method based on fast KNN algorithm. First, continuous data discretization processing and character data digitization processing are performed on the original data. Then the feature reduction algorithm based on mutual information is used to reduce the features of the preprocessed data. Finally, the fast KNN algorithm is used to classify and detect the feature-reduced data, and the classification results are output. The above method was verified on the KDD CUP99 data set and compared with existing technologies. The method proposed in this article clearly has the advantages of high classification efficiency and classification accuracy.

XGBoost odor prediction model: finding the structure-odor relationship of odorant molecules using the extreme gradient boosting algorithm.

Determining the structure-odor relationship has always been a very challenging task. The main challenge in investigating the correlation between the molecular structure and its associated odor is the ambiguous and obscure nature of verbally defined odor descriptors, particularly when the odorant molecules are from different sources. With the recent developments in machine learning (ML) technology, ML and data analytic techniques are significantly being used for quantitative structure-activity relationship (QSAR) in the chemistry domain toward knowledge discovery where the traditional Edisonian methods have not been useful. The smell perception of odorant molecules is one of the aforementioned tasks, as olfaction is one of the least understood senses as compared to other senses. In this study, the XGBoost odor prediction model was generated to classify smells of odorant molecules from their SMILES strings. We first collected the dataset of 1278 odorant molecules with seven basic odor descriptors, and then 1875 physicochemical properties of odorant molecules were calculated. To obtain relevant physicochemical features, a feature reduction algorithm called PCA was also employed. The ML model developed in this study was able to predict all seven basic smells with high precision (>99%) and high sensitivity (>99%) when tested on an independent test dataset. The results of the proposed study were also compared with three recently conducted studies. The results indicate that the XGBoost-PCA model performed better than the other models for predicting common odor descriptors. The methodology and ML model developed in this study may be helpful in understanding the structure-odor relationship.Communicated by Ramaswamy H. Sarma.

Study of the Immediately Detection of Mild Traumatic Brain Injury by Feature Engineering on Electroencephalography.

The electroencephalographic (EEG) diagnosis of mild traumatic brain injury (mTBI) is not usually timely, and the detection is often performed several hours or days after the trauma, leading to a decrease in the accuracy of its detection. In this study, EEG signals are recorded immediately after mTBI by connecting a bipolar single lead to injured animals. And three types of EEG features, namely time domain, frequency domain, and nonlinear dynamics, are screened for optimal feature subset in mTBI detection. First, EEG signals of animals are recorded before and after establishing the animal model of mTBI. Second, signal preprocessing, feature extraction, and feature preprocessing are performed to obtain the full-feature dataset, and 1442 feature subsets are obtained by 15 feature reduction algorithms extracted from combinations of 47 features. Ultimately, the support vector machines and K-nearest neighbor algorithms are trained and tested respectively, and their performance is comprehensively compared to determine the optimal feature subset for mTBI detection. In the EEG dataset collected in this study, a total of eight feature subsets extracted from combinations of original 47 features and classification models with 100% accuracy are obtained. This study shows the perspective of immediately detecting mTBI based on a bipolar single-lead EEG.

Parallel Selector for Feature Reduction

In the field of rough set, feature reduction is a hot topic. Up to now, to better guide the explorations of this topic, various devices regarding feature reduction have been developed. Nevertheless, some challenges regarding these devices should not be ignored: (1) the viewpoint provided by a fixed measure is underabundant; (2) the final reduct based on single constraint is sometimes powerless to data perturbation; (3) the efficiency in deriving the final reduct is inferior. In this study, to improve the effectiveness and efficiency of feature reduction algorithms, a novel framework named parallel selector for feature reduction is reported. Firstly, the granularity of raw features is quantitatively characterized. Secondly, based on these granularity values, the raw features are sorted. Thirdly, the reordered features are evaluated again. Finally, following these two evaluations, the reordered features are divided into groups, and the features satisfying given constraints are parallel selected. Our framework can not only guide a relatively stable feature sequencing if data perturbation occurs but can also reduce time consumption for feature reduction. The experimental results over 25 UCI data sets with four different ratios of noisy labels demonstrated the superiority of our framework through a comparison with eight state-of-the-art algorithms.

Automated human emotion recognition using hybrid approach based on sensitivity analysis on residual time-frequency plane with online learning algorithm

Emotions are neurophysiological changes that are modulated by central and peripheral nerve modulators that cause transient agitation due to the synchronized firing of nerve cells. Hence, the electroencephalogram (EEG) signal perfectly reflected the stimulated excitations of emotions. The excitation of these neurons corresponding to different emotions varies with instantaneous time. The focus of this work is to develop effective, efficient, and scalable data mining tools to study brain signals. The sensitivity analysis technique is employed to extract the valuable knowledge embedded in the vast amounts of brain data. It measures dilation parameters as a function of spectral domain changes without compromising the integrity of the original data. A residual time-frequency wavelet analogy is introduced to explore the EEG signals in the high-dimensional plane. Various nonlinear statistics are measured to extract the underlying complexity and variance discrepancy from the residual time-frequency domain. The feature matrix is further processed by a feature reduction and selection algorithm followed by a classifier to maximize the classification probability. This emotionally labeled feature matrix is augmented to enhance the dimension of the feature space. A bi-LSTM based deep learning network is used to secure a maximum probability of classification in a short time. The results reveal that the proposed algorithm maintains multiple potentials to achieve an excellent classification accuracy of order 78.37%, which helps researchers to understand the EEG patterns associated with each emotion.

Bacterial Colony Phenotyping with Hyperspectral Elastic Light Scattering Patterns.

The elastic light-scatter (ELS) technique, which detects and discriminates microbial organisms based on the light-scatter pattern of their colonies, has demonstrated excellent classification accuracy in pathogen screening tasks. The implementation of the multispectral approach has brought further advantages and motivated the design and validation of a hyperspectral elastic light-scatter phenotyping instrument (HESPI). The newly developed instrument consists of a supercontinuum (SC) laser and an acousto-optic tunable filter (AOTF). The use of these two components provided a broad spectrum of excitation light and a rapid selection of the wavelength of interest, which enables the collection of multiple spectral patterns for each colony instead of relying on single band analysis. The performance was validated by classifying microflora of green-leafed vegetables using the hyperspectral ELS patterns of the bacterial colonies. The accuracy ranged from 88.7% to 93.2% when the classification was performed with the scattering pattern created at a wavelength within the 473-709 nm region. When all of the hyperspectral ELS patterns were used, owing to the vastly increased size of the data, feature reduction and selection algorithms were utilized to enhance the robustness and ultimately lessen the complexity of the data collection. A new classification model with the feature reduction process improved the overall classification rate to 95.9%.

A multi-objective based radiomics feature selection method for response prediction following radiotherapy

Objective. Radiomics contains a large amount of mineable information extracted from medical images, which has important significance in treatment response prediction for personalized treatment. Radiomics analyses generally involve high dimensions and redundant features, feature selection is essential for construction of prediction models. Approach. We proposed a novel multi-objective based radiomics feature selection method (MRMOPSO), where the number of features, sensitivity, and specificity are jointly considered as optimization objectives in feature selection. The MRMOPSO innovated in the following three aspects: (1) Fisher score to initialize the population to speed up the convergence; (2) Min-redundancy particle generation operations to reduce the redundancy between radiomics features, a truncation strategy was introduced to further reduce the number of features effectively; (3) Particle selection operations guided by elitism strategies to improve local search ability of the algorithm. We evaluated the effectiveness of the MRMOPSO by using a multi-institution oropharyngeal cancer dataset from The Cancer Imaging Archive. 357 patients were used for model training and cross validation, an additional 64 patients were used for evaluation. Main results. The area under the curve (AUC) of our method achieved AUCs of 0.82 and 0.84 for cross validation and independent dataset, respectively. Compared with classical feature selection methods, the AUC of MRMOPSO is significantly higher than the Lasso (AUC = 0.74, p-value = 0.02), minimal-redundancy-maximal-relevance criterion (mRMR) (AUC = 0.73, p-value = 0.05), F-score (AUC = 0.48, p-value < 0.01), and mutual information (AUC = 0.69, p-value < 0.01) methods. Compared to single-objective methods, the AUC of MRMOPSO is 12% higher than those of the genetic algorithm (GA) (AUC = 0.68, p-value = 0.02) and particle swarm optimization algorithm (AUC = 0.72, p-value = 0.05) methods. Compared to other multi-objective feature selection methods, the AUC of MRMOPSO is 14% higher than those of multiple objective particle swarm optimization (MOPSO) (AUC = 0.68, p-value = 0.02) and nondominated sorting genetic algorithm II (NSGA2) (AUC = 0.70, p-value = 0.03). Significance. We proposed a multi-objective based radiomics feature selection method. Compared to conventional feature reduction algorithms, the proposed algorithm effectively reduced feature dimension, and achieved superior performance, with improved sensitivity and specificity, for response prediction in radiotherapy.

EEG-based Affect Classification with Machine Learning Algorithms

In this paper, we aim to study the EEG-based emotion recognition problem. First, we use clustering algorithm to determine the target class of emotions and perform binary classification of emotion along its arousal and valence dimension. Then we compare two different feature extraction methods, i.e., wavelet transform (resulting in wavelet-based features) and nonlinear dynamics analysis (leading to features of approximate entropy and sample entropy). Five feature reduction algorithms are compared in terms of emotion classification accuracy. Furthermore, four types of machine learning classifiers, including k-nearest neighbor (KNN), naive bayes (NB), support vector machine (SVM) and random forest (RF), are also compared. The results on the DEAP physiological data show that the combination of kernel spectral regression (KSR) and random forest leads to the best binary classification of emotions and that the EEG gamma rhythm is closely correlated to variations in emotions.

Electroencephalography signals and neurodevelopment after Kawasaki disease: A pilot study.

Acute Kawasaki disease (KD) induces central nervous system inflammation and excessive irritability. Long-term impacts on children's neurodevelopment have only been studied marginally. This pilot study aimed to describe the neuropsychological profile of primary school-aged children with a history of KD and to explore the impacts of KD on electroencephalography (EEG) markers associated with attention and brain maturation. Fifteen children (8.8 ± 2.5 years) were recruited 4.9 ± 2.7 years after KD onset. Intellectual abilities, long-term memory, and auditory sustained attention were evaluated. Parents completed standardized questionnaires assessing (1) executive functioning; (2) internalizing and externalizing difficulties; (3) attention deficit hyperactivity disorder symptoms; and (4) autism spectrum disorder symptoms. Theta/beta ratio (TBR) and alpha peak (AP) were extracted from resting-state EEG and compared with 32 controls (8.9 ± 2.1 years). The alpha band was analyzed using a feature reduction algorithm to detect potential groupings. Performances showed preserved intellectual abilities and memory. Sustained attention performance was within the lower range for 4/14 participants (29%), with considerable parental reports of inattention (43%), working memory difficulties (50%), and hyperactivity-impulsivity (36%). No alterations in the TBR were found but the KD group presented a significantly lower AP amplitude ratio. A clear separation of KD cohort into two clusters showed that acute irritability is associated with a weaker AP. Despite overall preserved cognitive functions, there is a possible association between KD and attention deficit concerns. This first EEG-based study indicates alpha peak abnormality after KD, predominantly in children with acute irritability. Longitudinal studies are warranted to better characterize patients' neurodevelopmental trajectory.

TSFNFR: Two-stage fuzzy neighborhood-based feature reduction with binary whale optimization algorithm for imbalanced data classification

Recently developed imbalanced data classification models are mainly focused on the majority class samples. In addition, several whale optimization-based feature reduction models are inefficient for high-dimensional data, readily fall into the local optimum, and are subject to difficulties associated with achieving a global optimal feature subset, due to high costs. To overcome the drawbacks, in this study, a two-stage feature reduction model using fuzzy neighborhood rough sets (FNRS) and the binary whale optimization algorithm (BWOA) was developed for imbalanced data classification. First, to indicate the sample fuzziness of mixed data, a similarity measure between samples based on fuzzy neighborhood was defined to investigate the similarity matrix and fuzzy neighborhood granule, and a new FNRS model was presented by constructing lower and upper approximations. By considering the uneven distribution of classes, the boundary-based feature significance measure was developed to minimize the influence of the uncertainty in boundary region for imbalanced data. Second, fuzzy neighborhood roughness and decision entropy were investigated based on FNRS, and by integrating these above measures, fuzzy neighborhood decision entropy was proposed to evaluate the fuzziness and roughness of the fuzzy neighborhood for imbalanced data. The external and internal significance metrics were proposed to achieve the preselected feature subset in the first stage. Third, in this second stage, a new control factor was defined to control the position of whales, and a novel fitness function was developed to evaluate the selected feature subset from imbalanced datasets. Thereafter, the immune regulation strategy of artificial immune was introduced into the BWOA to design the mixed selection probability, to divide the whale population. Two local interference strategies were applied to adjust the whale position and prevent BWOA trapped in the local optimum. Thus, an optimal feature subset was achieved by constantly iterating the BWOA. Finally, a two-stage feature reduction algorithm was designed to handle imbalanced and high-dimensional data, where the particle swarm optimization (PSO) algorithm was employed to determine the different optimized parameters for this two-stage algorithm. Experiments conducted on 22 datasets revealed that the proposed algorithm is efficient for two-class and multiclass datasets.

Integration of feature vectors from raw laboratory, medication and procedure names improves the precision and recall of models to predict postoperative mortality and acute kidney injury

Manuscripts that have successfully used machine learning (ML) to predict a variety of perioperative outcomes often use only a limited number of features selected by a clinician. We hypothesized that techniques leveraging a broad set of features for patient laboratory results, medications, and the surgical procedure name would improve performance as compared to a more limited set of features chosen by clinicians. Feature vectors for laboratory results included 702 features total derived from 39 laboratory tests, medications consisted of a binary flag for 126 commonly used medications, procedure name used the Word2Vec package for create a vector of length 100. Nine models were trained: baseline features, one for each of the three types of data Baseline + Each data type, (all features, and then all features with feature reduction algorithm. Across both outcomes the models that contained all features (model 8) (Mortality ROC-AUC 94.32 ± 1.01, PR-AUC 36.80 ± 5.10 AKI ROC-AUC 92.45 ± 0.64, PR-AUC 76.22 ± 1.95) was superior to models with only subsets of features. Featurization techniques leveraging a broad away of clinical data can improve performance of perioperative prediction models.

Applying multidate Sentinel-2 data for forest-type classification in complex broadleaf forest stands

AbstractBiodiversity assessment and forest management require accurate tree species maps, which can be provided by remote sensing. Whereas the application of high-spatial resolution remote sensing data is constrained by high costs, Sentinel-2 (S2) satellites provide free imagery with appropriate spatial, spectral and temporal resolutions for mapping of various forest traits across larger spatial scales. Here we assessed the potential of multidate S2 as well as a Digital Elevation Model (DEM) in classifying tree species across a highly structured and heterogeneous broadleaf forest ecosystem in the Hyrcanian zone of northern Iran. We applied multidate S2 and DEM data as input to a variable selection using random forests algorithm for feature reduction. Ten forest types were classified using random forest algorithm and to evaluate the results we computed area-adjusted confusion matrices. Classifications based on single-date S2 data reached overall accuracies of 67–74 per cent, whereas results for multidate S2 images increased the accuracy by ~28 per cent. Joint use of DEM data along with multidate S2 images showed improvement of overall accuracy by ~3 per cent. In addition, we studied the effect of topographic correction of S2 data on classification performance. The results imply that applying topographically corrected imagery had no significant effect on the classification accuracy. Our results demonstrate the high potential of freely available multisource remotely sensed data for broadleaf tree species classification across complex broad-leaved forest landscapes.

Examining stability of machine learning methods for predicting dementia at early phases of the disease

Dementia is a neuropsychiatric brain disorder that usually occurs when one or more brain cells stop working partially or at all. Diagnosis of this disorder in the early phases of the disease is a vital task to rescue patients’ lives from bad consequences and provide them with better healthcare. Machine learning methods have been proven to be accurate in predicting dementia in the early phases of the disease. The prediction of dementia depends heavily on the type of collected data which usually are gathered from Normalized Whole Brain Volume (nWBV) and Atlas Scaling Factor (ASF) which are normally measured and corrected from Magnetic Resonance Imaging (MRIs). Other biological features such as age and gender can also help in the diagnosis of dementia. Although many studies use machine learning for predicting dementia, we could not reach a conclusion on the stability of these methods for which one is more accurate under different experimental conditions. Therefore, this paper investigates the conclusion stability regarding the performance of machine learning algorithms for dementia prediction. To accomplish this, a large number of experiments were run using 7 machine learning algorithms and two feature reduction algorithms namely, Information Gain (IG) and Principal Component Analysis (PCA). To examine the stability of these algorithms, thresholds of feature selection were changed for the IG from 20% to 100% and the PCA dimension from 2 to 8. This has resulted in 7×9 + 7×7= 112 experiments. In each experiment, various classification evaluation data were recorded. The obtained results show that among seven algorithms the support vector machine and Naïve Bayes are the most stable algorithms while changing the selection threshold. Also, it was found that using IG would seem more efficient than using PCA for predicting Dementia. These promising results open the door to a new era of early prognosis of Alzheimer’s Disease and Related Dementias (ADRD).

Sentiment Analysis Using Pre-Trained Language Model With No Fine-Tuning and Less Resource

Sentiment analysis has become popular when Natural Language Processing algorithms were proven to be able to process complex sentences with good accuracy. Recently, pre-trained language models such as BERT and mBERT, have been shown to be effective for improving language tasks. Most of the work in implementing the models focuses on fine-tuning BERT to achieve desirable results. However, this approach is resource-intensive and requires a long training time, up to a few hours on a GPU, depending on the dataset. Hence, this paper proposes a less complex system with less training time using the BERT model without the fine-tuning process and adopting a feature reduction algorithm to reduce sentence embeddings. The experimental results show that with 50% fewer sentence embeddings, the proposed system improves the accuracy by 1-2% with 71% less training time and 89% less memory usage. The proposed approach has also been proven to work for multilingual tasks by using a single mBERT model.

A competent multimodal recognition using imperfect region based face and gait cues using Median-LBPF and Median-LBPG based PCA followed by LDA

Today, all conventional computer based multimodal biometric system for human recognition that combines face and gait cues have primarily focused on recognition from perfect face and gait images. There are circumstances in which perfect face and gait images may not be obtainable which means probe images are imperfect. This paper proposes new methods Median Local Binary Pattern of Face image (Median-LBPF) and Gait image (Median-LBPG) to extract the features of imperfect face and gait images efficiently for better recognition. Initially the given imperfect face and gait images are divided into six overlapped regions called top, bottom, left, right, vertical center, horizontally center overlapped half images. The features of these six overlapped regions of imperfect face and gait images in the spatial domain are extracted by using Median-LBPF and Median-LBPG. Subsequently the dimensionality of the feature sets are reduced by a two stage feature reduction algorithms Principal Component Analysis (PCA) followed by Linear Discriminant Analysis (LDA). Next for classification, Euclidean distance measure is used to calculate the minimum of minimum distance between the six overlapped regions of given imperfect face and gait probe images and the corresponding regions of all six overlapped regions in the training sets separately called face decision and gait decision. Finally the face decision and gait decision are fused at decision level for recognition. The proposed methods are tested by using publically available data sets ORL face and CASIA gait. The experimental results show that features of a region of face and gait images are adequate for recognition and its average recognition performance is same as perfect face and gait images.

Hierarchical Boosting Dual-Stage Feature Reduction Ensemble Model for Parkinson's Disease Speech Data.

As a neurodegenerative disease, Parkinson’s disease (PD) is hard to identify at the early stage, while using speech data to build a machine learning diagnosis model has proved effective in its early diagnosis. However, speech data show high degrees of redundancy, repetition, and unnecessary noise, which influence the accuracy of diagnosis results. Although feature reduction (FR) could alleviate this issue, the traditional FR is one-sided (traditional feature extraction could construct high-quality features without feature preference, while traditional feature selection could achieve feature preference but could not construct high-quality features). To address this issue, the Hierarchical Boosting Dual-Stage Feature Reduction Ensemble Model (HBD-SFREM) is proposed in this paper. The major contributions of HBD-SFREM are as follows: (1) The instance space of the deep hierarchy is built by an iterative deep extraction mechanism. (2) The manifold features extraction method embeds the nearest neighbor feature preference method to form the dual-stage feature reduction pair. (3) The dual-stage feature reduction pair is iteratively performed by the AdaBoost mechanism to obtain instances features with higher quality, thus achieving a substantial improvement in model recognition accuracy. (4) The deep hierarchy instance space is integrated into the original instance space to improve the generalization of the algorithm. Three PD speech datasets and a self-collected dataset are used to test HBD-SFREM in this paper. Compared with other FR algorithms and deep learning algorithms, the accuracy of HBD-SFREM in PD speech recognition is improved significantly and would not be affected by a small sample dataset. Thus, HBD-SFREM could give a reference for other related studies.