Ensemble feature selection for high dimensional data: a new method and a comparative study

The curse of dimensionality is based on the fact that high dimensional data is often difficult to work with. A large number of features can increase the noise of the data and thus the error of a learning algorithm. Feature selection is a solution for such problems where there is a need to reduce the data dimensionality. Different feature selection algorithms may yield feature subsets that can be considered local optima in the space of feature subsets. Ensemble feature selection combines independent feature subsets and might give a better approximation to the optimal subset of features. We propose an ensemble feature selection approach based on feature selectors’ reliability assessment. It aims at providing a unique and stable feature selection without ignoring the predictive accuracy aspect. A classification algorithm is used as an evaluator to assign a confidence to features selected by ensemble members based on their associated classification performance. We compare our proposed approach to several existing techniques and to individual feature selection algorithms. Results show that our approach often improves classification performance and feature selection stability for high dimensional data sets.     

Consistent tuning parameter selection in high dimensional sparse linear regression

An exhaustive search as required for traditional variable selection methods is impractical in high dimensional statistical modeling. Thus, to conduct variable selection, various forms of penalized estimators with good statistical and computational properties, have been proposed during the past two decades. The attractive properties of these shrinkage and selection estimators, however, depend critically on the size of regularization which controls model complexity. In this paper, we consider the problem of consistent tuning parameter selection in high dimensional sparse linear regression where the dimension of the predictor vector is larger than the size of the sample. First, we propose a family of high dimensional Bayesian Information Criteria (HBIC), and then investigate the selection consistency, extending the results of the extended Bayesian Information Criterion (EBIC), in Chen and Chen (2008) to ultra-high dimensional situations. Second, we develop a two-step procedure, the SIS+AENET, to conduct variable selection in p>n situations. The consistency of tuning parameter selection is established under fairly mild technical conditions. Simulation studies are presented to confirm theoretical findings, and an empirical example is given to illustrate the use in the internet advertising data.     

Model-free conditional independence feature screening for ultrahigh dimensional data

Feature screening plays an important role in ultrahigh dimensional data analysis. This paper is concerned with conditional feature screening when one is interested in detecting the association between the response and ultrahigh dimensional predictors (e.g., genetic makers) given a low-dimensional exposure variable (such as clinical variables or environmental variables). To this end, we first propose a new index to measure conditional independence, and further develop a conditional screening procedure based on the newly proposed index. We systematically study the theoretical property of the proposed procedure and establish the sure screening and ranking consistency properties under some very mild conditions. The newly proposed screening procedure enjoys some appealing properties. (a) It is model-free in that its implementation does not require a specification on the model structure; (b) it is robust to heavy-tailed distributions or outliers in both directions of response and predictors; and (c) it can deal with both feature screening and the conditional screening in a unified way. We study the finite sample performance of the proposed procedure by Monte Carlo simulations and further illustrate the proposed method through two real data examples.     

A selective overview of feature screening for ultrahigh-dimensional data

High-dimensional data have frequently been collected in many scientific areas including genomewide association study, biomedical imaging, tomography, tumor classifications, and finance. Analysis of highdimensional data poses many challenges for statisticians. Feature selection and variable selection are fundamental for high-dimensional data analysis. The sparsity principle, which assumes that only a small number of predictors contribute to the response, is frequently adopted and deemed useful in the analysis of high-dimensional data.Following this general principle, a large number of variable selection approaches via penalized least squares or likelihood have been developed in the recent literature to estimate a sparse model and select significant variables simultaneously. While the penalized variable selection methods have been successfully applied in many highdimensional analyses, modern applications in areas such as genomics and proteomics push the dimensionality of data to an even larger scale, where the dimension of data may grow exponentially with the sample size. This has been called ultrahigh-dimensional data in the literature. This work aims to present a selective overview of feature screening procedures for ultrahigh-dimensional data. We focus on insights into how to construct marginal utilities for feature screening on specific models and motivation for the need of model-free feature screening procedures.     

Genetic programming for feature construction and selection in classification on high-dimensional data

Classification on high-dimensional data with thousands to tens of thousands of dimensions is a challenging task due to the high dimensionality and the quality of the feature set. The problem can be addressed by using feature selection to choose only informative features or feature construction to create new high-level features. Genetic programming (GP) using a tree-based representation can be used for both feature construction and implicit feature selection. This work presents a comprehensive study to investigate the use of GP for feature construction and selection on high-dimensional classification problems. Different combinations of the constructed and/or selected features are tested and compared on seven high-dimensional gene expression problems, and different classification algorithms are used to evaluate their performance. The results show that the constructed and/or selected feature sets can significantly reduce the dimensionality and maintain or even increase the classification accuracy in most cases. The cases with overfitting occurred are analysed via the distribution of features. Further analysis is also performed to show why the constructed feature can achieve promising classification performance.     

����ά��������ģ�͵�PGFR����ɸѡ

In this paper, we consider the ultra-high dimensional partially linear model, where the dimensionality p of linear component is much larger than the sample size n, and p can be as large as an exponential of the sample size n. Firstly, we transform the ultra-high dimensional partially linear model into the ultra-high dimensional linear model based the profile technique used in the semiparametric regression. Secondly, in order to finish the variable screening for high-dimensional linear component, we propose a variable screening method called as the profile greedy forward regression (PGFR) by combining the greedy algorithm with the forward regression (FR) method. The proposed PGFR method not only considers the correlation between the covariates, but also identifies all relevant predictors consistently and possesses the screening consistency property under the some regularity conditions. We further propose the BIC criterion to determine whether the selected model contains the true model with probability tending to one. Finally, some simulation studies and a real application are conducted to examine the finite sample performance of the proposed PGFR procedure.     

Sure feature screening for high-dimensional dichotomous classification

The curse of high-dimensionality has emerged in the statistical fields more and more frequently. Many techniques have been developed to address this challenge for classification problems. We propose a novel feature screening procedure for dichotomous response data. This new method can be implemented as easily as t-test marginal screening approach, and the proposed procedure is free of any subexponential tail probability conditions and moment requirement and not restricted in a specific model structure. We prove that our method possesses the sure screening property and also illustrate the effect of screening by Monte Carlo simulation and apply it to a real data example.     

Near-optimal feature selection for large databases

We analyse a new optimization-based approach for feature selection that uses the nested partitions method for combinatorial optimization as a heuristic search procedure to identify good feature subsets. In particular, we show how to improve the performance of the nested partitions method using random sampling of instances. The new approach uses a two-stage sampling scheme that determines the required sample size to guarantee convergence to a near-optimal solution. This approach therefore also has attractive theoretical characteristics. In particular, when the algorithm terminates in finite time, rigorous statements can be made concerning the quality of the final feature subset. Numerical results are reported to illustrate the key results, and show that the new approach is considerably faster than the original nested partitions method and other feature selection methods.     

Aggregating multiple classification results using fuzzy integration and stochastic feature selection

Classifying magnetic resonance spectra is often difficult due to the curse of dimensionality; scenarios in which a high-dimensional feature space is coupled with a small sample size. We present an aggregation strategy that combines predicted disease states from multiple classifiers using several fuzzy integration variants. Rather than using all input features for each classifier, these multiple classifiers are presented with different, randomly selected, subsets of the spectral features. Results from a set of detailed experiments using this strategy are carefully compared against classification performance benchmarks. We empirically demonstrate that the aggregated predictions are consistently superior to the corresponding prediction from the best individual classifier.     

Model-free feature screening for ultrahigh-dimensional data conditional on some variables

In this paper, the conditional distance correlation (CDC) is used as a measure of correlation to develop a conditional feature screening procedure given some significant variables for ultrahigh-dimensional data. The proposed procedure is model free and is called conditional distance correlation-sure independence screening (CDC-SIS for short). That is, we do not specify any model structure between the response and the predictors, which is appealing in some practical problems of ultrahigh-dimensional data analysis. The sure screening property of the CDC-SIS is proved and a simulation study was conducted to evaluate the finite sample performances. Real data analysis is used to illustrate the proposed method. The results indicate that CDC-SIS performs well.     

����ά���ݱ߼ʾ�����Ȼ����ɸѡ����

The additive model is a more flexible nonparametric statistical model which allows a data-analytic transform of the covariates.When the number of covariates is big and grows exponentially with the sample size the urgent issue is to reduce dimensionality from high to a moderate scale. In this paper, we propose and investigate marginal empirical likelihood screening methods in ultra-high dimensional additive models. The proposed nonparametric screening method selects variables by ranking a measure of the marginal empirical likelihood ratio evaluated at zero to differentiate contributions of each covariate given to a response variable. We show that, under some mild technical conditions, the proposed marginal empirical likelihood screening methods have a sure screening property and the extent to which the dimensionality can be reduced is also explicitly quantified. We also propose a data-driven thresholding and an iterative marginal empirical likelihood methods to enhance the finite sample performance for fitting sparse additive models. Simulation results and real data analysis demonstrate the proposed methods work competitively and performs better than competitive methods in error of a heteroscedastic case.     

A discrete particle swarm optimization method for feature selection in binary classification problems

This paper investigates the feature subset selection problem for the binary classification problem using logistic regression model. We developed a modified discrete particle swarm optimization (PSO) algorithm for the feature subset selection problem. This approach embodies an adaptive feature selection procedure which dynamically accounts for the relevance and dependence of the features included the feature subset. We compare the proposed methodology with the tabu search and scatter search algorithms using publicly available datasets. The results show that the proposed discrete PSO algorithm is competitive in terms of both classification accuracy and computational performance.     

On the k-sample Behrens-Fisher problem for high-dimensional data

For several decades, much attention has been paid to the two-sample Behrens-Fisher (BF) problem which tests the equality of the means or mean vectors of two normal populations with unequal variance/covariance structures. Little work, however, has been done for the k-sample BF problem for high dimensional data which tests the equality of the mean vectors of several high-dimensional normal populations with unequal covariance structures. In this paper we study this challenging problem via extending the famous Scheffe’s transformation method, which reduces the k-sample BF problem to a one-sample problem. The induced one-sample problem can be easily tested by the classical Hotelling’s T ² test when the size of the resulting sample is very large relative to its dimensionality. For high dimensional data, however, the dimensionality of the resulting sample is often very large, and even much larger than its sample size, which makes the classical Hotelling’s T ² test not powerful or not even well defined. To overcome this difficulty, we propose and study an L ²-norm based test. The asymptotic powers of the proposed L ²-norm based test and Hotelling’s T ² test are derived and theoretically compared. Methods for implementing the L ²-norm based test are described. Simulation studies are conducted to compare the L ²-norm based test and Hotelling’s T ² test when the latter can be well defined, and to compare the proposed implementation methods for the L ²-norm based test otherwise. The methodologies are motivated and illustrated by a real data example. The work was supported by the National University of Singapore Academic Research Grant (Grant No. R-155-000-085-112)     

Heuristics for feature selection in mathematical programming discriminant analysis models

In developing a classification model for assigning observations of unknown class to one of a number of specified classes using the values of a set of features associated with each observation, it is often desirable to base the classifier on a limited number of features. Mathematical programming discriminant analysis methods for developing classification models can be extended for feature selection. Classification accuracy can be used as the feature selection criterion by using a mixed integer programming (MIP) model in which a binary variable is associated with each training sample observation, but the binary variable requirements limit the size of problems to which this approach can be applied. Heuristic feature selection methods for problems with large numbers of observations are developed in this paper. These heuristic procedures, which are based on the MIP model for maximizing classification accuracy, are then applied to three credit scoring data sets.     

A novel approach to estimation of E. coli promoter gene sequences: Combining feature selection and least square support vector machine (FS_LSSVM)

In this paper, we have investigated the real-world task of recognizing biological concepts in DNA sequences. Recognizing promoters in strings that represent nucleotides (one of A, G, T, or C) has been performed using a novel approach based on combining feature selection (FS) and least square support vector machine (LSSVM). Dimensionality of Escherichia coli promoter gene sequences dataset has 57 attributes and 106 samples including 53 promoters and 53 non-promoters. The proposed system consists of two parts. Firstly, we have used the FS process to reduce the dimensionality of E. coli promoter gene sequences dataset that has 57 attributes. So the dimensionality of this dataset has been reduced to 4 attributes by means of FS process.Secondly, LSSVM classifier algorithm has been run to estimation the E. coli promoter gene sequences. In order to show the performance of the proposed system, we have used the success rate, sensitivity and specificity analysis, 10-fold cross validation, and confusion matrix. Whilst only LSSVM classifier has been obtained 80% success rate using 10-fold cross validation, the proposed system has been obtained 100% success rate for same condition. These obtained results indicate that the proposed approach improve the success rate in recognizing promoters in strings that represent nucleotides.     

Variable selection in censored quantile regression with high dimensional data

We propose a two-step variable selection procedure for censored quantile regression with high dimensional predictors. To account for censoring data in high dimensional case, we employ effective dimension reduction and the ideas of informative subset idea. Under some regularity conditions, we show that our procedure enjoys the model selection consistency. Simulation study and real data analysis are conducted to evaluate the finite sample performance of the proposed approach.     

Overview of Feature Screening Methods for Ultra-high Dimensional Data

With the improvement of data collection and storage capacity, ultra-high dimensional data\ucite{9}, that is, dimensionality with the exponential growth of samples appears in many scientific neighborhoods. At this time, penalized variable selection methods generally encounter three challenges: computational expediency, statistical accuracy, and algorithmic stability, which are limited in handling ultra-high dimensional problems. Fan and Lv\ucite{9} proposed the method of ultra-high dimensional feature screening, and achieved a lot of research results in the past ten years, which has become the most popular field of research in statistics. This paper mainly introduces the related work of ultra-high dimensional screening method from four aspects: the screening methods with model hypothesis, including parametric, non-parametric and semi-parametric model hypothesis, model-free hypothesis, and screening methods for special data. Finally, we briefly discuss the existing problems of ultra-high dimensional screening methods and some future directions.     

Bridge estimation for generalized linear models with a diverging number of parameters

Variable selection is fundamental to high dimensional generalized linear models. A number of variable selection approaches have been proposed in the literature. This paper considers the problem of variable selection and estimation in generalized linear models via a bridge penalty in the situation where the number of parameters diverges with the sample size. Under reasonable conditions the consistency of the bridge estimator can be achieved. Furthermore, it can select the nonzero coefficients with a probability converging to 1 and the estimators of nonzero coefficients have the asymptotic normality, namely the oracle property. Our simulations indicate that the bridge penalty is an effective consistent model selection technique and is comparable to the smoothly clipped absolute deviation procedure. A real example analysis is presented.     

Evolutionary methods for unsupervised feature selection using Sammon’s stress function

In this paper, four methods are proposed for feature selection in an unsupervised manner by using genetic algorithms. The proposed methods do not use the class label information but select a set of features using a task independent criterion that can preserve the geometric structure (topology) of the original data in the reduced feature space. One of the components of the fitness function is Sammon’s stress function which tries to preserve the topology of the high dimensional data when reduced into the lower dimensional one. In this context, in addition to using a fitness criterion, we also explore the utility of unfitness criterion to select chromosomes for genetic operations. This ensures higher diversity in the population and helps unfit chromosomes to become more fit. We use four different ways for evaluation of the quality of the features selected: Sammon error, correlation between the inter-point distances in the two spaces, a measure of preservation of cluster structure found in the original and reduced spaces and a classifier performance. The proposed methods are tested on six real data sets with dimensionality varying between 9 and 60. The selected features are found to be excellent in terms of preservation topology (inter-point geometry), cluster structure and classifier performance. We do not compare our methods with other methods because, unlike other methods, using four different ways we check the quality of the selected features by finding how well the selected features preserve the “structure” of the original data.     

An exact feature selection algorithm based on rough set theory

Feature reduction based on rough set theory is an effective feature selection method in pattern recognition applications. Finding a minimal subset of the original features is inherent in rough set approach to feature selection. As feature reduction is a Nondeterministic Polynomial‐time‐hard problem, it is necessary to develop fast optimal or near‐optimal feature selection algorithms. This article aims to propose an exact feature selection algorithm in rough set that is efficient in terms of computation time. The proposed algorithm begins the examination of a solution tree by a breadth‐first strategy. The pruned nodes are held in a version of the trie data structure. Based on the monotonic property of dependency degree, all subsets of the pruned nodes cannot be optimal solutions. Thus, by detecting these subsets in trie, it is not necessary to calculate their dependency degree. The search on the tree continues until the optimal solution is found. This algorithm is improved by selecting an initial search level determined by the hill‐climbing method instead of searching the tree from the level below the root. The length of the minimal reduct and the size of data set can influence which starting search level is more efficient. The experimental results using some of the standard UCI data sets, demonstrate that the proposed algorithm is effective and efficient for data sets with more than 30 features. © 2014 Wiley Periodicals, Inc. Complexity 20: 50–62, 2015