Multi-class Data Research Articles

NOREVA: enhanced normalization and evaluation of time-course and multi-class metabolomic data.

Biological processes (like microbial growth & physiological response) are usually dynamic and require the monitoring of metabolic variation at different time-points. Moreover, there is clear shift from case-control (N=2) study to multi-class (N>2) problem in current metabolomics, which is crucial for revealing the mechanisms underlying certain physiological process, disease metastasis, etc. These time-course and multi-class metabolomics have attracted great attention, and data normalization is essential for removing unwanted biological/experimental variations in these studies. However, no tool (including NOREVA 1.0 focusing only on case-control studies) is available for effectively assessing the performance of normalization method on time-course/multi-class metabolomic data. Thus, NOREVA was updated to version 2.0 by (i) realizing normalization and evaluation of both time-course and multi-class metabolomic data, (ii) integrating 144 normalization methods of a recently proposed combination strategy and (iii) identifying the well-performing methods by comprehensively assessing the largest set of normalizations (168 in total, significantly larger than those 24 in NOREVA 1.0). The significance of this update was extensively validated by case studies on benchmark datasets. All in all, NOREVA 2.0 is distinguished for its capability in identifying well-performing normalization method(s) for time-course and multi-class metabolomics, which makes it an indispensable complement to other available tools. NOREVA can be accessed at https://idrblab.org/noreva/.

Towards Cross-Modality Medical Image Segmentation with Online Mutual Knowledge Distillation

The success of deep convolutional neural networks is partially attributed to the massive amount of annotated training data. However, in practice, medical data annotations are usually expensive and time-consuming to be obtained. Considering multi-modality data with the same anatomic structures are widely available in clinic routine, in this paper, we aim to exploit the prior knowledge (e.g., shape priors) learned from one modality (aka., assistant modality) to improve the segmentation performance on another modality (aka., target modality) to make up annotation scarcity. To alleviate the learning difficulties caused by modality-specific appearance discrepancy, we first present an Image Alignment Module (IAM) to narrow the appearance gap between assistant and target modality data. We then propose a novel Mutual Knowledge Distillation (MKD) scheme to thoroughly exploit the modality-shared knowledge to facilitate the target-modality segmentation. To be specific, we formulate our framework as an integration of two individual segmentors. Each segmentor not only explicitly extracts one modality knowledge from corresponding annotations, but also implicitly explores another modality knowledge from its counterpart in mutual-guided manner. The ensemble of two segmentors would further integrate the knowledge from both modalities and generate reliable segmentation results on target modality. Experimental results on the public multi-class cardiac segmentation data, i.e., MM-WHS 2017, show that our method achieves large improvements on CT segmentation by utilizing additional MRI data and outperforms other state-of-the-art multi-modality learning methods.

The Comparison of LightGBM and XGBoost Coupling Factor Analysis and Prediagnosis of Acute Liver Failure

This paper focuses on the comparison of dimensionality reduction effect between LightGBM and XGBoost-FA. With respect to XGBoost, LightGBM can be built in the effect of dimensionality reduction via both Gradient-based One-Side Sampling(GOSS) and Exclusive Feature Bundling(EFB) algorithms, while XGBoost coupling with traditional dimensionality reduction tool Factor Analysis (XGBoost-FA) may also have dimensionality reduction effect. To present the empirical comparison, the prediagnosis dataset for the 2018 Kaggle competition Acute Liver Failure has been chosen as the research object. And pairwise comparison has been conducted among XGBoost, LightGBM, XGBoost-FA and LightGBM-FA. Concerning the test set, the vector (accuracy, log loss function, training time) of the above first four prediagnostic models are (0.75014, 0.569707, 10.5s), (0.75811, 0.576059,15.1s), (0.67786,0.663924,5.7s) and (0.67274,0.676019, 4.1s) respectively. It’s been found that the training time of XGBoost-FA (external dimensionality reduction) is shorter than that of LightGBM (build-in dimensionality reduction). Considering (accuracy, training time) being (0.82, 3.1s) published on Kaggle, the algorithm (logogram as K2a) is better than the four XGBoost-FA and LightGBM in both training time and accuracy. However, K2a removes more than 50% samples with missing values and only performs binary classification. For multi-class classification or data with a large number of missing values, XGBoost-FA is more suggested if higher operational time is required, while LightGBM is preferred if higher predictive accuracy is required. With XGBoost-FA or LightGBM being employed in AI medical services, doctors are more productive in diagnosis and treatment due to much more data support and less workload. Both complement each other.

A Non-Negative Matrix Factorization-Based Framework for the Analysis of Multi-Class Time-Series Single-Cell RNA-Seq Data

The development of single-cell RNA sequencing (scRNA-seq) has enabled gene expression to be quantified at single-cell resolution. Such advancement is expected to solve important issues that bulk RNA sequencing could not fully answer, such as inferring cell population heterogeneity, genetic variability of cells, detecting rare cell types, accurately predicting cell states and their localization. However, analyzing such large scale data, especially when they are sampled at multiple time points, brings new challenges in data mining informative genes, compared to single snapshot samples. It becomes even more complicated when gene expression patterns are to be mined from time-series scRNA-seq datasets generated from multiple conditions, which will constitute a data with gene, condition and time dimensions. Here, we focused on detecting gene expression patterns that well capture the underlying biological differences between time-series scRNA-seq datasets of three different types of stem cells. The gene expression profile of 2,128 time-series scRNA-seq samples from long-term hematopoietic stem cells (LT-HSC) and two of its progenitor cell types were analyzed using our framework. We have successfully detected condition specific feature genes that were able to achieve 90.03% classification accuracy between the three cell types. Investigating the genes and clusters detected by our framework, we found that cell cycle related genes showed significantly high variance between the three cell types. Such results and transcriptomic characters detected from our analysis were consistent with the original study. Collectively, the framework was able to successfully detect biological meaningful gene sets and expression patterns from multi-condition time-series scRNA-seq samples.

An automatic estimation of the ridge parameter for extreme learning machine

Extreme learning machine (ELM) is an emerging learning method with a single-hidden layer feed-forward neural network that involves obtaining a solution to the system of linear equations. Unlike traditional gradient-based back-propagating neural networks, ELM is computationally efficient with fast training speed and good generalization capability. However, most of the time when applied to real-time problems, the linear system becomes ill-posed in the structure and needs the inclusion of a ridge parameter to obtain a reliable solution, and hence, the selection of the ridge parameter (C) is a crucial task. The ridge parameter is chosen heuristically from a predefined set. The generalized cross-validation is a widely used technique for the automatic estimation of the same, which is computationally expensive as it involves inversion of large matrices. The focus of the proposed work is on pragmatic aspects of the time-efficient automatic estimation of ridge parameter that result in a better generalization performance. In this work, methods are proposed that use the L-curve and U-curve techniques to automatically estimate the ridge parameter, and these methods are effective in the estimation of the ridge parameter even for systems with larger data. Through extensive numerical results, it is shown that the proposed methods outperform the existing ones in terms of accuracy, precision, sensitivity, specificity, F1-score, and computational time on various benchmark binary as well as multiclass classification data sets. Finally, the proposed methods are statistically analyzed using the nonparametric Friedman ranking test, which is also proving the effectiveness of the proposed method as it is providing a better rank for the same over existing methods.

Machine learning-based cognitive impairment classification with optimal combination of neuropsychological tests.

IntroductionAn extensive battery of neuropsychological tests is currently used to classify individuals as healthy (HV), mild cognitively impaired (MCI), and with Alzheimer's disease (AD). We used machine learning models for effective cognitive impairment classification and optimized the number of tests for expeditious and inexpensive implementation.MethodsUsing random forests (RF) and support vector machine, we classified cognitive impairment in multi‐class data sets from Rush Religious Orders Study Memory and Aging Project, and National Alzheimer's Coordinating Center. We applied Fisher's linear discrimination and assessed importance of each test iteratively for feature selection.ResultsRF has best accuracy with increased sensitivity, specificity in this first ever multi‐class classification of HV, MCI, and AD. Moreover, a subset of six to eight tests shows equivalent classification accuracy as an entire battery of tests.DiscussionsFully automated feature selection approach reveals six to eight tests comprising episodic, semantic memory, perceptual orientation, and executive functioning can accurately classify the cognitive status, ensuring minimal subject burden.

Combining One-vs-One Decomposition and Instance-Based Learning for Multi-Class Classification

Multi-class classification is one of the most important supervised learning problems and can be solved by either designing direct multi-class classifiers (direct strategy) or decomposing it into a set of binary classification problems (indirect strategy). Direct strategy only needs training one unified classifier while indirect strategy, especially the one-vs-one decomposition method, has shown its superiority and been utilized by some popular software packages. In this article, a first attempt towards bridging the gap between direct strategy and one-vs-one decomposition for multi-class classification is conducted, and accordingly a novel approach named CODIL is proposed. Specifically, CODIL firstly transforms the class vector into a ternary label matrix (only with {-1, 0, +1}) via one-vs-one rule, where each column of the label matrix corresponds to a pair of classes. Then, CODIL determines the binary label vector (only with {-1, +1}) for unseen instance by exploiting the manifold structure information residing in its k nearest neighbors, where each element of the label vector denotes the unseen instance's prediction for its corresponding class pair. Finally, the class for unseen instance is returned via majority voting based on the binary label vector. Extensive comparative studies are conducted between CODIL and six well-established multi-class approaches over seventeen benchmark multi-class data sets. The experimental results show the superiority of the proposed CODIL approach against the compared approaches.

Class-Incremental Learning of Convolutional Neural Networks Based on Double Consolidation Mechanism

Class-incremental learning is a model learning technique that can help classification models incrementally learn about new target classes and realize knowledge accumulation. It has become one of the major concerns of the machine learning and classification community. To overcome the catastrophic forgetting that occurs when the network is trained sequentially on a multi-class data stream, a double consolidation class-incremental learning (DCCIL) method is proposed. In the incremental learning process, the network parameters are adjusted by combining knowledge distillation and elastic weight consolidation, so that the network can better maintain the recognition ability of the old classes while learning the new ones. The incremental learning experiment is designed, and the proposed method is compared with the popular incremental learning methods such as EWC, LwF, and iCaRL. Experimental results show that the proposed DCCIL method can achieve better incremental accuracy than that of the current popular incremental learning algorithms, which can effectively improve the expansibility and intelligence of the classification model.

Random Balance ensembles for multiclass imbalance learning

Random Balance strategy (RandBal) has been recently proposed for constructing classifier ensembles for imbalanced, two-class data sets. In RandBal, each base classifier is trained with a sample of the data with a random class prevalence, independent of the a priori distribution. Hence, for each sample, one of the classes will be undersampled while the other will be oversampled. RandBal can be applied on its own or can be combined with any other ensemble method. One particularly successful variant is RandBalBoost which integrates Random Balance and boosting. Encouraged by the success of RandBal, this work proposes two approaches which extend RandBal to multiclass imbalance problems. Multiclass imbalance implies that at least two classes have substantially different proportion of instances. In the first approach proposed here, termed Multiple Random Balance (MultiRandBal), we deal with all classes simultaneously. The training data for each base classifier are sampled with random class proportions. The second approach we propose decomposes the multiclass problem into two-class problems using one-vs-one or one-vs-all, and builds an ensemble of RandBal ensembles. We call the two versions of the second approach OVO-RandBal and OVA-RandBal, respectively. These two approaches were chosen because they are the most straightforward extensions of RandBal for multiple classes. Our main objective is to evaluate both approaches for multiclass imbalanced problems. To this end, an experiment was carried out with 52 multiclass data sets. The results suggest that both MultiRandBal, and OVO/OVA-RandBal are viable extensions of the original two-class RandBal. Collectively, they consistently outperform acclaimed state-of-the art methods for multiclass imbalanced problems.

A Hardware-Deployable Neuromorphic Solution for Encoding and Classification of Electronic Nose Data.

In several application domains, electronic nose systems employing conventional data processing approaches incur substantial power and computational costs and limitations, such as significant latency and poor accuracy for classification. Recent developments in spike-based bio-inspired approaches have delivered solutions for the highly accurate classification of multivariate sensor data with minimized computational and power requirements. Although these methods have addressed issues related to efficient data processing and classification accuracy, other areas, such as reducing the processing latency to support real-time application and deploying spike-based solutions on supported hardware, have yet to be studied in detail. Through this investigation, we proposed a spiking neural network (SNN)-based classifier, implemented in a chip-emulation-based development environment, that can be seamlessly deployed on a neuromorphic system-on-a-chip (NSoC). Under three different scenarios of increasing complexity, the SNN was determined to be able to classify real-valued sensor data with greater than 90% accuracy and with a maximum latency of 3 s on the software-based platform. Highlights of this work included the design and implementation of a novel encoder for artificial olfactory systems, implementation of unsupervised spike-timing-dependent plasticity (STDP) for learning, and a foundational study on early classification capability using the SNN-based classifier.

Multiple Factors Act Differently in Decision Making in the East Asian Region: Assessing Methods of Self-Construal Using Classification Tree Analysis

Cultural resemblance between Japan and Korea has been assumed, because both East Asian cultures are generally characterized by collectivism. However, subsequent investigations suggest a contrast between the two cultures in the decision-making process, concurrent with a growing call for a more elaborate model of self-construal. This study compares the diverse ways of self-construal underlying decisions regarding invitations between Japan and Korea. We conducted a scenario-based questionnaire in which 377 Japanese and Korean college students made decisions regarding whether to invite someone to an activity, to do the activity alone, or to not do the activity. Utilizing classification tree modeling to analyze multivariate categorical data, we examined the participants’ selected actions according to two factors: interpersonal relations (i.e., whether the other was a friend or an unfamiliar but favored classmate) and the locus of interest in a given situation (i.e., whether the event was of interest to the speaker, hearer, or both parties). The results revealed cultural variations in invitations. Japanese students seem to be interdependent, in that they hesitate to invite others unless they have an assured interpersonal relationship, and tend to wait until they are sure of the other’s interest, even when interested themselves. In contrast, Korean students appear more independent in considering the other’s interest and pursuing their own interest. This study highlights the complex ways of self-construal within the alleged collectivist East Asian region, and also demonstrates the applicability of classification tree analysis in differentiating rank orders of multiple factors influencing people’s decision making across cultures.

Multiclass Data Imbalance Oversampling Techniques (Mudiot) and Random Selection of Features

Class imbalance is a serious issue in classification problem. If a class is unevenly distributed the classification algorithm unable to classify the response variable, which will result in inaccuracy. The technique Multiclass Data Imbalance Oversampling Techniques (MuDIOT) is to find out the factors which have a hidden negative impact on classification. To alleviate the negative impact the technique MuDIOT concentrates on balancing the data and the result minimizes the problems raised due to uneven distribution of classes. The dataset chosen has a multiclass distribution problem and it is handled to produce better results of classification.

Cost-sensitive stacked auto-encoders for intrusion detection in the Internet of Things

Intrusion Detection System (IDS) is an important tool for protecting the Internet of Things (IoT) networks against cyber-attacks. Traditional IDSs can only distinguish between normal and abnormal behaviors. On the other hand, modern techniques can identify the kind of attack so that the appropriate reactions can be carried out against each type of attack. However, these techniques always suffer from the class-imbalance which affects the performance of IDS. In this paper, we propose a cost-sensitive stacked auto-encoder, CSSAE, to deal with class imbalance problem in IDS. CSSAE generates a cost matrix in which a unique cost is assigned to each class based on the distribution of different classes. This matrix is created in the first stage of CSSAE. In the second phase, a two-layer stacked auto-encoder is applied to learn features with better distinguish between the minority and the majority classes. These costs are used in the feature learning of deep learning, where the parameters of the neural network are modified by applying the corresponding costs in the cost function layer. The proposed method is able to perform on both binary-class data and multiclass data. Two well-known KDD CUP 99 and NSL-KDD datasets are used to evaluate the performance of CSSAE. Compared with other IDSs that have not considered class imbalance problem, CSSAE shows better performance in the detection of low-frequency attacks.

Extreme Learning Machine Weight Optimization using Particle Swarm Optimization to Identify Sugar Cane Disease

Sugar cane disease is a major factor in reducing sugar cane yields. The low intensity of experts to go into the field to check the condition of sugar cane causes the handling of sugarcane disease tends to be slow. This problem can be solved by instilling expert intelligence on sugar cane into an expert system. In this study the method of classification of sugar cane disease was proposed using Extreme Learning Machine (ELM). However, ELM alone is not enough to classify multilabel and multiclass disease case data in this study. Therefore, it is proposed to optimize the weight of hidden neurons in ELM using Particle Swarm Optimization (PSO). The experimental results show that the classification using ELM alone can reach an accuracy rate of 71%. After the weight of hidden neurons from ELM was optimized, the accuracy rate became 79.92% or an increase of 8.92%.

Logical analysis of multiclass data with relaxed patterns

An efficient and robust algorithm based on mixed integer linear programming is proposed to extend the Logical Analysis of Data (LAD) methodology to solve multiclass classification problems, where One-vs-Rest learning models are constructed to classify observations in predefined classes. The proposed algorithm uses two control parameters, homogeneity and prevalence, for identifying relaxed (fuzzy) patterns in multiclass datasets. The utility of the proposed method is demonstrated through experiments on multiclass benchmark datasets. Numerical experiments show that the efficiency and performance of the proposed multiclass LAD method with relaxed patterns is comparable to, if not better than, those of the previously developed LAD based multiclass classification as well as other well-known supervised learning methods.

Area-Proportional Visualization for Circular Data

Data visualization is important for statistical analysis, as it helps convey information efficiently and shed lights on the hidden patterns behind data in a visual context. It is particularly helpful to display circular data in a two-dimensional space to accommodate its nonlinear support space and reveal the underlying circular structure which is otherwise not obvious in one-dimension. In this article, we first formally categorize circular plots into two types, either height- or area-proportional, and then describe a new general methodology that can be used to produce circular plots, particularly in the area-proportional manner, which in our opinion is the more appropriate choice. Formulas are given that are fairly simple yet effective to produce various circular plots, such as smooth density curves, histograms, rose diagrams, dot plots, and plots for multiclass data. Supplemental materials for this article are available online.

Collaborative and geometric multi-kernel learning for multi-class classification

The multi-class classification is the problem of classifying the sample into one of three or more classes. In this paper, we propose an algorithm named collaborative and geometric multi-kernel learning (CGMKL) to classify multi-class data into corresponding class directly. The CGMKL uses the Multiple Empirical Kernel Learning (MEKL) to map the sample into multiple kernel spaces, and then trains the softmax function in each kernel space. To realize the collaborative learning, one regularization term, which controls the consistent outputs of samples in different kernel spaces, provides the complementary information. Moreover, another regularization term exhibits the classification result with a geometric feature by reducing the within-class distance of the outputs of samples. Extensive Experiments on the multi-class data sets validate the effectiveness of the CGMKL.

Multi-class twitter data categorization and geocoding with a novel computing framework

This study details the progress in transportation data analysis with a novel computing framework in keeping with the continuous evolution of the computing technology. The computing framework combines the Labeled Latent Dirichlet Allocation (L-LDA)-incorporated Support Vector Machine (SVM) classifier with the supporting computing strategy on publicly available Twitter data in determining transportation-related events to provide reliable information to travelers. The analytical approach includes analyzing tweets using text classification and geocoding locations based on string similarity. A case study conducted for the New York City and its surrounding areas demonstrates the feasibility of the analytical approach. Approximately 700,010 tweets are analyzed to extract relevant transportation-related information for one week. The SVM classifier achieves >85% accuracy in identifying transportation-related tweets from structured data. To further categorize the transportation-related tweets into sub-classes: incident, congestion, construction, special events, and other events, three supervised classifiers are used: L-LDA, SVM, and L-LDA incorporated SVM. Findings from this study demonstrate that the analytical framework, which uses the L-LDA incorporated SVM, can classify roadway transportation-related data from Twitter with over 98.3% accuracy, which is significantly higher than the accuracies achieved by standalone L-LDA and SVM.

Optimizing shapelets quality measure for imbalanced time series classification

Time series classification has been considered as one of the most challenging problems in data mining and is widely used in a broad range of fields. A biased distribution leads to classification on minority time series objects more severe. A commonly taken approach is to extract or select the representative features to retain the structure of a time series object. However, when the data distribution is imbalanced, the traditional features cannot represent time series effectively, especially in multi-class environment. In this paper, Shapelets — a primitive time series mining technology — is applied to extract the most representative subsequences. Especially, we verify that IG (Information Gain) is unsuitable as a shapelet quality measure for imbalanced data sets. Nevertheless, we propose two quality measures for shapelets on imbalanced binary and multi-class problem respectively. Based on extracted shapelet features, we select the diversified top-k shapelets based on new quality measure to represent the top-k best features and achieve this procedure on map-reduce framework. Lastly, two oversampling methods based on shapelet features are proposed to re-balance the binary and multi-class time series data sets. We validated our methods on the benchmark data sets by comparing with the canonical classifiers and the state-of-the-art time series algorithms. It is verified that the proposed algorithms perform more competitive than the compared methods in statistical significance.

FTIR spectroscopy of freeze-dried human sera as a novel approach for dengue diagnosis

Attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy was used as a diagnostic tool to identify the biochemical markers expressed differentially in dengue infected and healthy freeze-dried human blood sera. A total of 77 sera samples; 57 DENV infected and 20 healthy were processed. Multivariate data classification algorithms; principal component analysis (PCA), linear discriminant analysis (LDA), and partial least square regression (PLSR) were used for classified differentiation between selected wave number regions of FTIR spectra of both types of samples. PCA-LDA model applied on our data showed sensitivity and specificity of 89% and 95% respectively. PLSR model successfully predicted the infection signs based on biochemical changes in serum samples with an accuracy of R2 = 0.9980. This study provides an initial insight that ATR-FTIR spectroscopy coupled with multivariate data classification algorithms can be used as an effective tool for dengue diagnosis. This is the first report of the use of freeze-dried human blood sera for FTIR analysis.