Multiclass Logistic Regression Research Articles

Dependence of farmers' livelihoods on environmental resource in key ecological function area: A case study of Gannan Plateau, China

The key ecological function area plays an important role in providing ecosystem service. As the main players of economic activities in this area, the excessive dependence of local farmers on environmental resource will seriously influence ecological environment quality, and thus impair the ecosystem service. At present, it is urgent to explore the dependence of farmers' livelihoods on environmental resource and its influencing factors in key ecological function areas, which would provide reference for the formulation of environmental friendly regional development policies. We selected the Gannan Plateau as study area. Based on 581 survey data, we analyzed the dependence of farmers' livelihood on environmental resource, and used the ordinal multi-class Logistic regression model to identify its key impact factors. The results showed that local farmers were highly dependent on environmental resource in Gannan Plateau, which was mainly reflected in three aspects of food self-sufficiency, daily energy consumption and income source, with the dependence of 57.3%, 56.9% and 37.4%, respectively. There were significant differences in the dependence of different types of farmers' livelihoods on environmental resource, in that farmers with higher levels of education and higher non-agricultural levels had lower dependence and the farmers with high dependency ratio were more dependent on environmental resources. The factors including family dependency ratio, labor education level, proportion of migrant workers in labor force, family member leadership, ecological policy, and altitude had significant impact on farmers' environmental resource dependence. Among them, family dependency ratio and altitude had positive impact on it, while the proportion of migrant workers in labor force, labor education level, family member leadership, ecological policy had negative impact.

Stochastic Tverberg Theorems With Applications in Multiclass Logistic Regression, Separability, and Centerpoints of Data

We present new stochastic geometry theorems that give bounds on the probability that $m$ random data classes all contain a point in common in their convex hulls. These theorems relate to the existe...

Constructing a Heterogeneous Training Dataset for Emotion Classification

Abstract Emotion classification deals with identifying emotions expressed within a text. Social media is generating a vast amount of emotion rich data in the form of tweets, status updates, blog posts etc. Tweets are a good representative of emotions a person usually expresses publicly. By analyzing the emotions in these tweets, one can get an idea of how a person feels about the subject they are referring to. Machine Learning (ML) techniques are widely used for analyzing emotions within the tweets. However, there are no balanced training datasets that can be used for training the ML classifiers. As a result, supervised classifiers demonstrate a poor performance with classifying emotions within texts particularly within the tweets. In addition to that, none of the available datasets are useful for training the classifiers to identify and classify emotions within the tweets. Therefore, in this paper we have proposed a novel approach for constructing a balanced heterogeneous training dataset for emotion classification of the tweets. Using the lexicon-based NRC classifier we have classified the textual instances in to four different emotions such as joyful, sad, angry and surprise. Using this as a training dataset we have trained six different machine learning models including the Multiclass Logistic Regression (MLR), Multinomial Naive Bayes (MLB), Random Forest (RF), Support Vector Machine (SVM), Convolutional Neural Network (CNN), and Recurrent Neural Network (RNN). Our study reveals that this approach has the potentiality in boosting the performance of the supervised classifiers for emotion classification within the tweets.

Multi-Factor Analysis of Single-Center Asthma Control in Xiamen, China.

We evaluated the effects of air pollutants, age, allergic history, family allergic history, treatment, treatment steps, and compliance on uncontrolled childhood asthma in Xiamen, China. The clinical data of children with asthma in the pediatric outpatient department of the First Affiliated Hospital of Xiamen University from January 2016 to June 2018 were analyzed retrospectively. According to the assessment of the patients' outcome including well-controlled, partly-controlled and uncontrolled, 7,211 cases of 3,268 patients were selected. Rank sum test and ordered multi-class logistic regression analysis were used. In the rank sum test, age, allergic history, family allergic history, season, treatment, treatment steps and compliance were found associated with uncontrolled rate (all P < 0.001). Logistic regression analysis showed that PM10, NO2, and SO2 raised uncontrolled-asthma rate (aOR 1.311, aOR 1.281, aOR 1.252, respectively). Older children had lower uncontrolled rate (OR = 0.849, 95% CI: 0.758–0.950), children with higher treatment steps had higher uncontrolled rate (OR = 1.227, 95%CI: 1.040–1.448), and children with better treatment compliance have lower uncontrolled rate (OR = 0.374 95% CI: 0.331–0.424). The order of the uncontrolled rate of asthma from high to low was winter, spring, autumn, and summer. PM10, NO2, SO2, age, season, treatment steps, and treatment compliance have significance for predicting the control rate of childhood asthma in Xiamen, China.

Exploiting the Potential of VGI Metadata to Develop A Data-Driven Framework for Predicting User’s Proficiency in OpenStreetMap Context

Volunteered geographic information (VGI) encourages citizens to contribute geographic data voluntarily that helps to enhance geospatial databases. VGI’s significant limitations are trustworthiness and reliability concerning data quality due to the anonymity of data contributors. We propose a data-driven model to address these issues on OpenStreetMap (OSM), a particular case of VGI in recent times. This research examines the hypothesis of evaluating the proficiency of the contributor to assess the credibility of the data contributed. The proposed framework consists of two phases, namely, an exploratory data analysis phase and a learning phase. The former explores OSM data history to perform feature selection, resulting in “OSM Metadata” summarized using principal component analysis. The latter combines unsupervised and supervised learning through K-means for user-clustering and multi-class logistic regression for user classification. We identified five major classes representing user-proficiency levels based on contribution behavior in this study. We tested the framework with India OSM data history, where 17% of users are key contributors, and 27% are unexperienced local users. The results for classifying new users are satisfactory with 95.5% accuracy. Our conclusions recognize the potential of OSM metadata to illustrate the user’s contribution behavior without the knowledge of the user’s profile information.

Changes of psychological stress and predictive factors in elderly patients with stroke after 6 months of discharge from rehabilitation

Objective To study the changes and predictors of psychological stress in elderly patients with acute stroke after 6 months of discharge. Methods Randomly selected patients who were admitted to the hospital from July 2017 to December 2017 were diagnosed as acutely stroke-free patients. Baseline variables were recorded for each patient at admission. After discharge from the hospital, the patient's perceptual pressure scale, the Healthy Stroke Scale, and the Depression Scale were recorded in the form of a follow-up visit to the patient at 2 weeks (T1), 3 months (T2), and 6 months (T3). 15 ratings. One-way ANOVA of the values of the T1 and T2, T1 and T3, T2 and T3 pressure gauges were performed. The healthy stroke scale score measured at the T1 time point, the functional independent measurement scale score, the depression scale score as the independent variable, and the perceived stress scale score at the T3 time point as the dependent variable, and the independent variable and the dependent variable. Disordered multi-class logistic regression analysis. Results After discharge, the scales of the two-week (T1), three-month (T2), and half-year (T3) time points were compared. The results showed that the health stroke scale scores at different time points (F=2.34), functional independent measurement scale scores (F=3.42), the depression scale score -15 (F=2.89), the perceived stress scale score-10 (F=4.57) was statistically different (P<0.05). There were statistical differences between the 2 weeks after discharge (T1) and 3 months after discharge (T2), 2 weeks after discharge (T1) and the second half of discharge (T3). There was a statistical difference (95% CI 1.56-7.20, 3.24-8.50, P<0.05). The results of Logistic regression analysis indicated that the functional independence score of patients with acute stroke after 2 weeks of discharge had a negative impact on psychological stress, while the depression score had a positive effect on psychological stress. Conclusions The psychological stress of surviving patients after acute stroke decreased at 3 months and 6 months. The functional independent measurement scale score and the depression scale score 2 weeks after discharge were psychological stress after discharge for 6 months. Independent influence factors. Key words: Depression; Stress; Stroke

Behavior Evaluation Based on Electroencephalograph and Personality in a Simulated Driving Experiment.

Assessments and predictions of driving behavior are very important to improve traffic safety. We hypothesized that there were some patterns of driving behaviors, and these patterns had some correlation with cognitive states and personalities. To test this hypothesis, an evaluation of driving status, based on electroencephalography (EEG) and steering behavior in a simulated driving experiment, was designed and performed. Unity 3D was utilized to design the simulated driving scene. A photoelectric encoder fixed on the steering wheel and the corresponding data collection, transmission, and storage device was developed by Arduino, to acquire the rotation direction, angle, angular velocity, and angular acceleration of the steering wheel. Biopac MP 150 was utilized to collect the EEG data simultaneously during driving. A total of 23 subjects (mean age 23.6 ± 1.3 years, driving years: 2.4 ± 1.6 years, 21 males and two females) participated in this study. The Fuzzy C-means algorithm (FCMA) was utilized to extract patterns of driving behavior and the cognitive state within the window width of 20 s. The behaviors were divided into five kinds, i.e., negative, normal, alert, stress, and violent behavior, respectively, based on the standard deviation of steering wheel data. The cognitive states were divided into four kinds, i.e., negative, calm, alert, and tension, respectively, based on the EEG data. The correlation of these data, together with the personality traits evaluated using Cattell 16 Personality Factor Questionnaire (16PF) were analyzed using multiclass logistic regression. Results indicated the significance of the cognitive state and seven personality traits [apprehension (O), rule consciousness (G), reasoning (B), emotional stability (C), liveliness (F), vigilance (L), and perfectionism (Q3)] in predicting driving behaviors, and the prediction accuracy was 80.2%. The negative and alert cognitive states were highly correlated with dangerous driving, including negative and violent behaviors. Personality traits complicate the relationship with driving behaviors, which may vary across different types of subjects and traffic accidents.

Study of satisfaction and influencing factors of medical insurance for urban residents′ in Taiyuan city

Objective To learn the satisfaction and influencing factors of medical insurance for urban residents in Taiyuan city. Methods The multi-stage random sampling method was used to survey the satisfaction of medical insurance for urban residents so covered in 4 districts of Taiyuan city.The questionnaire covered basic personal information, and comments of the residents on the three dimensions of medical insurance regarding the stakeholders.The data solicited were subject to descriptive analysis, Kruskal-Wallis test and ordered multi-class logistic regression analysis. Results The rate of overall satisfaction of urban residents′ medical insurance in Taiyuan city was 44.6%, that of average satisfaction was 48.3%, and that of dissatisfaction was 7.1%.For residents of different family size and physical health status, the differences of their satisfaction were statistically different(P 1). Conclusions The rate of overall satisfaction of medical insurance by urban residents in Taiyuan city is low.To improve their satisfaction, it is important to increase the proportion of reimbursement, expand the scope of reimbursement for diagnosis and treatment along with drugs, simplify the reimbursement procedures, and improve the service attitude of the service agencies. Key words: Urban residents; Medical insurance; Satisfactory; Influencing factors; Taiyuan city

Machine learning based prediction of metal hydrides for hydrogen storage, part II: Prediction of material class

The openly available dataset on hydrogen storage materials provided by the US Department of Energy was used to predict the optimal materials class of metal hydrides based on the desired properties, which included hydrogen-weight percent, heat of formation and operating temperature and pressure. We performed correlation and statistical analyses to investigate the statistical characteristics of each numeric features. We employed four classification algorithms: multiclass logistic regression, multiclass decision forest, multiclass decision jungle and multiclass neural network. Feature importance analysis was carried out to investigate how each classifier utilises the information available in the dataset. In overall, multiclass neural network classifier had better classification performance obtaining an accuracy of 0.80. The results suggest that the complex material class, followed by Mg is applicable for the most wide range of operating temperatures. Positive correlation was found between hydrogen weight percent, heat of formation and temperature, suggesting that the maximum hydrogen weight percent would be achieved in the complex material class operated at a high temperature.

Development of Predictive Models in Patients with Epiphora Using Lacrimal Scintigraphy and Machine Learning.

We developed predictive models using different programming languages and different computing platforms for machine learning (ML) and deep learning (DL) that classify clinical diagnoses in patients with epiphora. We evaluated the diagnostic performance of these models. Between January 2016 and September 2017, 250 patients with epiphora who underwent dacryocystography (DCG) and lacrimal scintigraphy (LS) were included in the study. We developed five different predictive models using ML tools, Python-based TensorFlow, R, and Microsoft Azure Machine Learning Studio (MAMLS). A total of 27 clinical characteristics and parameters including variables related to epiphora (VE) and variables related to dacryocystography (VDCG) were used as input data. Apart from this, we developed two predictive convolutional neural network (CNN) models for diagnosing LS images. We conducted this study using supervised learning. Among 500 eyes of 250 patients, 59 eyes had anatomical obstruction, 338 eyes had functional obstruction, and the remaining 103 eyes were normal. For the data set that excluded VE and VDCG, the test accuracies in Python-based TensorFlow, R, multiclass logistic regression in MAMLS, multiclass neural network in MAMLS, and nuclear medicine physician were 81.70%, 80.60%, 81.70%, 73.10%, and 80.60%, respectively. The test accuracies of CNN models in three-class classification diagnosis and binary classification diagnosis were 72.00% and 77.42%, respectively. ML-based predictive models using different programming languages and different computing platforms were useful for classifying clinical diagnoses in patients with epiphora and were similar to a clinician's diagnostic ability.

RAFFIA: Short-term Forest Fire Danger Rating Prediction via Multiclass Logistic Regression

Forest fire prevention is important because of human communities near forests or in the wildland-urban interfaces. Short-term forest fire danger rating prediction is an effective way to provide early guidance for forest fire managers. It can therefore effectively protect the forest resources and enhance the sustainability of the forest ecosystem. However, relevant existing forest fire danger rating prediction models operate well only when applied to distinct climates and fuel types separately. There are desires for an effective methodology, which can construct a specific short-term prediction model according to an evaluation of the data from that specific region. Moreover, a suitable method for prediction model construction needs to deal with some big data related computing challenges (i.e., data diversity coupled with complexity of solution space, and the requirement of real-time forest fire prevention application) when massively observed heterogeneous parameters are available for prediction (e.g., meteorology factor, the amount of litter in the area, soil moisture, etc.). To capture the influences of multiple prediction factors on the prediction results and effectively learn from fast cumulative historical big data, artificial intelligence methods are investigated in this paper, yielding a short-term Ratings of Forest Fire Danger Prediction via Multiclass Logistic Regression (or RAFFIA) model for forest fire danger rating online prediction. Experimental evaluations conducted on a sensor-based forest fire prevention experimental station show that RAFFIA (with 98.71% precision and 0.081 root mean square error) is more effective than the Least Square Fitting Regression (LSFR) and Random Forests (RF) prediction models.

Integrating Phosphoproteomics and Transcriptional Classifiers Reveals “Hidden Signaling” in Multiple Myeloma Including Differential KRAS and NRAS Mutant Effects

Introduction: Multiple myeloma (MM) is a complex disease that requires a sophisticated treatment strategy. Currently, no kinase inhibitors have been approved for MM despite their potential for supplementing current combination therapies. Previous functional studies have explored kinase dependency in MM by either a small molecule inhibitor library (Dhimolea et al. 2014 ASH) or RNA interference (Tiedemann et al. 2010, Blood 115:1594). However, owing to their off-target effects, these approaches are imprecise at dissecting signaling networks driving MM growth and survival. Here, we aim to improve diagnostic and prognostic measures and recommend small molecule-based treatments for MM patients by identifying vulnerable signaling patterns in disease using integrated transcriptome- and phosphoproteome-based predictive models.Methods: We employed two methods for measuring cellular signaling activity within a tumor sample. The first involves an unbiased phosphoproteome profiling of eight human myeloma cell lines (HMCL) in biological triplicates. Cells were lysed and digested with trypsin prior to enrichment for phosphorylated peptides by Fe3+-IMAC. Samples were analyzed on a Thermo Q-Exactive Plus mass spectrometer and data processed in MaxQuant to identify and quantify phosphorylation sites. To infer relative kinase activities, we applied kinase set enrichment analysis (KSEA). Drug screening was performed in 384-well plates with CellTiter-Glo as viability readout as previously described (Lam et al. 2018, Haematologica 103:1218). For transcriptome analysis, we implemented the gene expression-based signaling pathway prediction model called PROGENy (Schubert et al. 2018, Nat Comm. 9:20). We applied PROGENy to RNAseq data on 64 MM cell lines (www.keatslab.org) as well as plasma cells from >1000 patients in the MMRF CoMMpass study (research.themmrf.org). Furthermore, using our recently described approach (Way et al. 2018, Cell Rep. 23:172), we built a machine learning classifier to predict RAS genotype from the transcriptomic profiles of MM patients. A tenth of the data set was withheld for testing while the rest was used to train the multiclass logistic regression classifier with sparse penalty.Results: We measured the KSEA-predicted activities of 297 kinases across eight tested MM cell lines. Initially, we were surprised to find high predicted activity in the Ras signaling pathway for KRAS-codon 12 mutant cell lines but low predicted activity in NRAS-mutant cell lines (Fig. A: AMO1 harbors a non-canonical KRAS mutation at codon 146). We further explored this finding with our machine learning-based Ras classifier built on transcriptional data in the CoMMpass study. We identified 311, 405, and 390 genes whose expressions are characteristic of the WT RAS, KRAS mutant, and NRAS mutant genotype, respectively, with surprisingly limited overlap between KRAS and NRAS transcriptional signatures. Building on our KSEA analysis, we next performed a kinase inhibitor screen to evaluate the predictive value of the inferred kinase activities for drug sensitivity. Of 12 screened compounds, mTOR inhibitor INK128 displayed the strongest correlation between drug response and predicted kinase activity. Furthermore, we probed the potential of using pathway activity signatures as prognostic and therapeutic markers. To this end, we applied PROGENy to RNAseq data derived from CoMMpass patients and found that the MAPK signature stratifies patient survival with statistical significance, while the presence and absence of RAS mutations carry no prognostic value (Fig. B). Finally, by integrating RNAseq and drug screen data from the Cancer Dependency Map, we identified three compounds whose inhibitory effects strongly correlate with MAPK activity scores while no significant difference in drug sensitivity was detected between RAS WT and mutants.Conclusion: Both phosphoproteomics and a machine learning-based transcriptional classifier highlight a striking difference in the pattern of signaling between NRAS and KRAS mutants. In addition, we have demonstrated that PROGENy scores possess clinical value for prognostic and therapeutic use based on patient transcriptome data. Taken together, uncovering the cellular signaling networks dysregulated in MM may lead to improved precision medicine, particularly in stratifying patients who may benefit most from kinase inhibitor therapy. DisclosuresWiita:TeneoBio: Research Funding; Sutro Biopharma: Research Funding.

Correction to: Camera localization for a human-pose in 3D space using a single 2D human-pose image with landmarks: a multimedia social network emerging demand

Recovering a 3D human-pose in the form of an abstracted skeleton from a 2D image suffers from loss of depth information. Assuming the projected human-pose is represented by a set of 2D landmarks capturing the human-pose limbs, recovering back the original 3D locations is an ill posed problem. To recover a 3D configuration, camera localization in 3D space plays a major role, an inaccurate camera localization might mislead the recovery process. In this paper, we propose a 3D camera localization model using only human-pose appearance in a 2D image (i.e., the set of 2D landmarks). We apply a supervised multi-class logistic regression to assign the camera location in 3D space. In the learning process, we assume a set of predefined labeled camera locations. The features we train consist of relative length of limbs and 2D shape context. The goal is to build a relation between these projected landmarks and the camera location in 3D space. This kind of analysis allows us to reconstruct 3D human-poses based on the 2D projection only without any predefined camera parameters. Also, makes real-time multimedia exchange more reliable specially for human-pose related tasks. We test our model on a set of real images showing a variety of camera locations.

Willingness to pay for colorectal cancer screening in Guangzhou.

AIMTo measure the willingness to pay for colorectal cancer screening in Guangzhou, and to identify those factors associated with it.METHODSA face-to-face questionnaire survey for pre-screening population from free and non-free colonoscopy districts was used to collect information on demographic characteristics, health behaviours, the intention of the cancer screenings and willingness to pay for colorectal cancer screening. A total of 1243 participants who took part in the pre-screening for colorectal cancer in Guangzhou were collected in the study. Categorical data were compared using the χ2 test to analyse significant differences. Non-conditional logistic regression and multi-class logistic regression were also performed for multivariate analysis and to estimate the odds ratios.RESULTSThe percentage of participants willing to pay for colorectal cancer screening was 91.7%. “Unnecessary” was the dominant reason that participants gave for their unwillingness, accounting for 63.1%. Of those who were willing to pay, 29.2%, 20.7%, 14.8%, 13.0% and 22.4% of participants were willing to pay less than \\100, \\100-\\199, \\200-299, \\300-\\399 and more than \\400, respectively. Non-logistic regression analysis showed that respondents who were male, had a high level of education, were from the family with more children/older to raise, and accepted colorectal cancer screening were willing to pay for this screening. Multi-class logistic regression analysis showed that respondents with higher annual household income per capita, from government and private enterprises, government agency/institution and peasants, and less family medical expenditure were willing to pay more.CONCLUSIONWillingness to pay for colorectal cancer screening in Guangzhou is high, but the amount of willing to pay is not much.

Sentiment Analysis for Odd-Even Scheme in Delhi

Objectives: This paper analyzes odd-even traffic scheme using tweets posted on Twitter from December 2015 to August 2016. Twitter is a social network where users post their feelings, opinions and sentiments for any event using hashtags and mentions. The tweets posted publicly can be viewed by anyone interested. This paper transforms the unstructured tweets into structured information using open source libraries. Further objective is to build a model using machine learning classification techniques to classify unseen tweets on the same context. Methods/Analysis: This paper collects tweets on this event under hashtags. This study explores Dandelion Application Programming Interface for annotation of tweets for academic research. This paper uses machine learning classifications approaches for sentiment analysis and opinion mining. This paper presents empirical comparison of three supervised classification algorithms namely, Multinomial Naive Bayes, Support Vector Machines (SVM) and Multiclass Logistic Regression. The performances of these classifiers are evaluated through standard evaluation metrics. Findings: The experimental results reveal that SVM classifier outperforms the other two classification algorithms. This study may help in decision making of this event to some extent. Application: A large number of applications of sentiment and opinion mining can be designed using packages and freely open resources within a time frame now a days.

Sparse group lasso for multiclass functional logistic regression models

ABSTRACTSparsity-inducing penalties are useful tools for variable selection and are also effective for regression problems where the data are functions. We consider the problem of selecting not only variables but also decision boundaries in multiclass logistic regression models for functional data, using sparse regularization. The parameters of the functional logistic regression model are estimated in the framework of the penalized likelihood method with the sparse group lasso-type penalty, and then tuning parameters for the model are selected using the model selection criterion. The effectiveness of the proposed method is investigated through simulation studies and the analysis of a gene expression data set.

Early Destination Prediction with Spatio-temporal User Behavior Patterns

Predicting user behavior makes it possible to provide personalized services. Destination prediction (e.g. predicting a future location) can be applied to various practical applications. An example of destination prediction is personalized GIS services, which are expected to provide alternate routes to enable users to avoid congested roads. However, the destination prediction problem requires critical trade-offs between timing and accuracy. In this paper, we focus on early destination prediction as the central issue, as early recognition in destination prediction has not been fully explored. As an alternative to the traditional two basic approaches with trajectory tracking that narrow down the candidates with respect to the trip progress, and Next Place Prediction (NPP) that infers the future location of a user from user habits, we propose a new probabilistic model based on both conventional models. The advantage of our model is that it drastically narrows down the destination candidates efficiently at the early stage of a trip, owing to the staying information derived from the NPP approach. In other words, our approach achieves high prediction accuracy by considering both approaches at the same time. To implement our model, we employ SubSynE for state-of-the-art prediction based on trajectory tracking as well as a multi-class logistic regression based on user contexts. Despite the simplicity of our model, the proposed method provides improved performance compared to conventional approaches based on the experimental results using the GPS logs of 1,646 actual users from the commercial services.

A Multinomial Logistic Regression Approach for Arrhythmia Detection

Cardiovascular diseases are the leading causes on mortality in the world. Consequently, tools and methods providing useful and applicable insights into their assessment play a crucial role in the prediction and managements of specific heart conditions. In this article, we introduce a method based on multi-class Logistic Regression as a classifier to provide a powerful and accurate insight into cardiac arrhythmia, which is one of the predictors of serious vascular diseases. As suggested by our evaluation, this provides a robust, scalable, and accurate system, which can successfully tackle the challenges posed by the utilisation of big data in the medical sector.

Radiomic analysis reveals DCE-MRI features for prediction of molecular subtypes of breast cancer.

The purpose of this study was to investigate the role of features derived from breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and to incorporated clinical information to predict the molecular subtypes of breast cancer. In particular, 60 breast cancers with the following four molecular subtypes were analyzed: luminal A, luminal B, human epidermal growth factor receptor-2 (HER2)-over-expressing and basal-like. The breast region was segmented and the suspicious tumor was depicted on sequentially scanned MR images from each case. In total, 90 features were obtained, including 88 imaging features related to morphology and texture as well as dynamic features from tumor and background parenchymal enhancement (BPE) and 2 clinical information-based parameters, namely, age and menopausal status. An evolutionary algorithm was used to select an optimal subset of features for classification. Using these features, we trained a multi-class logistic regression classifier that calculated the area under the receiver operating characteristic curve (AUC). The results of a prediction model using 24 selected features showed high overall classification performance, with an AUC value of 0.869. The predictive model discriminated among the luminal A, luminal B, HER2 and basal-like subtypes, with AUC values of 0.867, 0.786, 0.888 and 0.923, respectively. An additional independent dataset with 36 patients was utilized to validate the results. A similar classification analysis of the validation dataset showed an AUC of 0.872 using 15 image features, 10 of which were identical to those from the first cohort. We identified clinical information and 3D imaging features from DCE-MRI as candidate biomarkers for discriminating among four molecular subtypes of breast cancer.

Sparse coding based dense feature representation model for hyperspectral image classification

We present a sparse coding based dense feature representation model (a preliminary version of the paper was presented at the SPIE Remote Sensing Conference, Dresden, Germany, 2013) for hyperspectral image (HSI) classification. The proposed method learns a new representation for each pixel in HSI through the following four steps: sub-band construction, dictionary learning, encoding, and feature selection. The new representation usually has a very high dimensionality requiring a large amount of computational resources. We applied the l 1 /l q regularized multiclass logistic regression technique to reduce the size of the new representation. We integrated the method with a linear support vector machine (SVM) and a composite kernels SVM (CKSVM) to discriminate different types of land cover. We evaluated the proposed algorithm on three well-known HSI datasets and compared our method to four recently developed classification methods: SVM, CKSVM, simultaneous orthogonal matching pursuit, and image fusion and recursive filtering. Experimental results show that the proposed method can achieve better overall and average classification accuracies with a much more compact representation leading to more efficient sparse models for HSI classification.