Iterative construction of hierarchical classifiers for phishing website detection

This article is devoted to a new iterative construction of hierarchical classifiers in SimpleCLI for the detection of phishing websites. Our new construction of hierarchical systems creates ensembles of ensembles in SimpleCLI by iteratively linking a top-level ensemble to another middle-level ensemble instead of a base classifier so that the top-level ensemble can generate a large multilevel system. This new construction makes it easy to set up and run such large systems in SimpleCLI. The present article concentrates on the investigation of performance of the iterative construction of such classifiers for the example of detection of phishing websites. We carried out systematic experiments evaluating several essential ensemble techniques as well as more recent approaches and studying their performance as parts of the iterative construction of hierarchical classifiers. The results presented here demonstrate that the iterative construction of hierarchical classifiers performed better than the base classifiers and standard ensembles. This example of application to the classification of phishing websites shows that the new iterative construction combining diverse ensemble techniques into the iterative construction of hierarchical classifiers can be applied to increase the performance in situations where data can be processed on a large computer. © 2014 ACADEMY PUBLISHER.

Enhancing medical named entity recognition with an extended segment representation technique

Objective : The objective of this paper is to formulate an extended segment representation (SR) technique to enhance named entity recognition (NER) in medical applications.

Methods : An extension to the IOBES (Inside/Outside/Begin/End/Single) SR technique is formulated. In the proposed extension, a new class is assigned to words that do not belong to a named entity (NE) in one context but appear as an NE in other contexts. Ambiguity in such cases can negatively affect the results of classification-based NER techniques. Assigning a separate class to words that can potentially cause ambiguity in NER allows a classifier to detect NEs more accurately; therefore increasing classification accuracy.

Results : The proposed SR technique is evaluated using the i2b2 2010 medical challenge data set with eight different classifiers. Each classifier is trained separately to extract three different medical NEs, namely treatment, problem, and test. From the three experimental results, the extended SR technique is able to improve the average F1-measure results pertaining to seven out of eight classifiers. The kNN classifier shows an average reduction of 0.18% across three experiments, while the C4.5 classifier records an average improvement of 9.33%.

EEG data classification using wavelet features selected by Wilcoxon statistics

This paper introduces a method to classify EEG signals using features extracted by an integration of wavelet transform and the nonparametric Wilcoxon test. Orthogonal Haar wavelet coefficients are ranked based on the Wilcoxon test’s statistics. The most prominent discriminant wavelets are assembled to form a feature set that serves as inputs to the naïve Bayes classifier. Two benchmark datasets, named Ia and Ib, downloaded from the brain–computer interface (BCI) competition II are employed for the experiments. Classification performance is evaluated using accuracy, mutual information, Gini coefficient and F-measure. Widely used classifiers, including feedforward neural network, support vector machine, k-nearest neighbours, ensemble learning Adaboost and adaptive neuro-fuzzy inference system, are also implemented for comparisons. The proposed combination of Haar wavelet features and naïve Bayes classifier considerably dominates the competitive classification approaches and outperforms the best performance on the Ia and Ib datasets reported in the BCI competition II. Application of naïve Bayes also provides a low computational cost approach that promotes the implementation of a potential real-time BCI system.

Automatic generation of meta classifiers with large levels for distributed computing and networking

This paper is devoted to a case study of a new construction of classifiers. These classifiers are called automatically generated multi-level meta classifiers, AGMLMC. The construction combines diverse meta classifiers in a new way to create a unified system. This original construction can be generated automatically producing classifiers with large levels. Different meta classifiers are incorporated as low-level integral parts of another meta classifier at the top level. It is intended for the distributed computing and networking. The AGMLMC classifiers are unified classifiers with many parts that can operate in parallel. This make it easy to adopt them in distributed applications. This paper introduces new construction of classifiers and undertakes an experimental study of their performance. We look at a case study of their effectiveness in the special case of the detection and filtering of phishing emails. This is a possible important application area for such large and distributed classification systems. Our experiments investigate the effectiveness of combining diverse meta classifiers into one AGMLMC classifier in the case study of detection and filtering of phishing emails. The results show that new classifiers with large levels achieved better performance compared to the base classifiers and simple meta classifiers classifiers. This demonstrates that the new technique can be applied to increase the performance if diverse meta classifiers are included in the system.

Classification ensemble to improve medical named entity recognition

An accurate Named Entity Recognition (NER) is important for knowledge discovery in text mining. This paper proposes an ensemble machine learning approach to recognise Named Entities (NEs) from unstructured and informal medical text. Specifically, Conditional Random Field (CRF) and Maximum Entropy (ME) classifiers are applied individually to the test data set from the i2b2 2010 medication challenge. Each classifier is trained using a different set of features. The first set focuses on the contextual features of the data, while the second concentrates on the linguistic features of each word. The results of the two classifiers are then combined. The proposed approach achieves an f-score of 81.8%, showing a considerable improvement over the results from CRF and ME classifiers individually which achieve f-scores of 76% and 66.3% for the same data set, respectively.

Android malware detection with contrasting permission patterns

As the risk of malware is sharply increasing in Android platform, Android malware detection has become an important research topic. Existing works have demonstrated that required permissions of Android applications are valuable for malware analysis, but how to exploit those permission patterns for malware detection remains an open issue. In this paper, we introduce the contrasting permission patterns to characterize the essential differences between malwares and clean applications from the permission aspect. Then a framework based on contrasting permission patterns is presented for Android malware detection. According to the proposed framework, an ensemble classifier, Enclamald, is further developed to detect whether an application is potentially malicious. Every contrasting permission pattern is acting as a weak classifier in Enclamald, and the weighted predictions of involved weak classifiers are aggregated to the final result. Experiments on real-world applications validate that the proposed Enclamald classifier outperforms commonly used classifiers for Android Malware Detection.

Prediction interval estimation for electricity price and demand using support vector machines

Uncertainty is known to be a concomitant factor of almost all the real world commodities such as oil prices, stock prices, sales and demand of products. As a consequence, forecasting problems are becoming more and more challenging and ridden with uncertainty. Such uncertainties are generally quantified by statistical tools such as prediction intervals (Pis). Pis quantify the uncertainty related to forecasts by estimating the ranges of the targeted quantities. Pis generated by traditional neural network based approaches are limited by high computational burden and impractical assumptions about the distribution of the data. A novel technique for constructing high quality Pis using support vector machines (SVMs) is being proposed in this paper. The proposed technique directly estimates the upper and lower bounds of the PI in a short time and without any assumptions about the data distribution. The SVM parameters are tuned using particle swarm optimization technique by minimization of a modified Pi-based objective function. Electricity price and demand data of the Ontario electricity market is used to validate the performance of the proposed technique. Several case studies for different months indicate the superior performance of the proposed method in terms of high quality PI generation and shorter computational times.

Intelligent facial emotion recognition using a layered encoding cascade optimization model

In this research, we propose a facial expression recognition system with a layered encoding cascade optimization model. Since generating an effective facial representation is a vital step to the success of facial emotion recognition, a modified Local Gabor Binary Pattern operator is first employed to derive a refined initial face representation and we then propose two evolutionary algorithms for feature optimization including (i) direct similarity and (ii) Pareto-based feature selection, under the layered cascade model. The direct similarity feature selection considers characteristics within the same emotion category that give the minimum within-class variation while the Pareto-based feature optimization focuses on features that best represent each expression category and at the same time provide the most distinctions to other expressions. Both a neural network and an ensemble classifier with weighted majority vote are implemented for the recognition of seven expressions based on the selected optimized features. The ensemble model also automatically updates itself with the most recent concepts in the data. Evaluated with the Cohn-Kanade database, our system achieves the best accuracies when the ensemble classifier is applied, and outperforms other research reported in the literature with 96.8% for direct similarity based optimization and 97.4% for the Pareto-based feature selection. Cross-database evaluation with frontal images from the MMI database has also been conducted to further prove system efficiency where it achieves 97.5% for Pareto-based approach and 90.7% for direct similarity-based feature selection and outperforms related research for MMI. When evaluated with 90° side-view images extracted from the videos of the MMI database, the system achieves superior performances with >80% accuracies for both optimization algorithms. Experiments with other weighting and meta-learning combination methods for the construction of ensembles are also explored with our proposed ensemble showing great adpativity to new test data stream for cross-database evaluation. In future work, we aim to incorporate other filtering techniques and evolutionary algorithms into the optimization models to further enhance the recognition performance.

Prediction interval estimation of electricity prices using PSO-tuned support vector machines

Uncertainty of the electricity prices makes the task of accurate forecasting quite difficult for the electricity market participants. Prediction intervals (PIs) are statistical tools which quantify the uncertainty related to forecasts by estimating the ranges of the future electricity prices. Traditional approaches based on neural networks (NNs) generate PIs at the cost of high computational burden and doubtful assumptions about data distributions. In this work, we propose a novel technique that is not plagued with the above limitations and it generates high-quality PIs in a short time. The proposed method directly generates the lower and upper bounds of the future electricity prices using support vector machines (SVM). Optimal model parameters are obtained by the minimization of a modified PI-based objective function using a particle swarm optimization (PSO) technique. The efficiency of the proposed method is illustrated using data from Ontario, Pennsylvania-New Jersey-Maryland (PJM) interconnection day-ahead and real-time markets.

Examining performance of aggregation algorithms for neural network-based electricity demand forecasting

The aim of this research is to examine the efficiency of different aggregation algorithms to the forecasts obtained from individual neural network (NN) models in an ensemble. In this study an ensemble of 100 NN models are constructed with a heterogeneous architecture. The outputs from NN models are combined by three different aggregation algorithms. These aggregation algorithms comprise of a simple average, trimmed mean, and a Bayesian model averaging. These methods are utilized with certain modifications and are employed on the forecasts obtained from all individual NN models. The output of the aggregation algorithms is analyzed and compared with the individual NN models used in NN ensemble and with a Naive approach. Thirty-minutes interval electricity demand data from Australian Energy Market Operator (AEMO) and the New York Independent System Operator's web site (NYISO) are used in the empirical analysis. It is observed that the aggregation algorithm perform better than many of the individual NN models. In comparison with the Naive approach, the aggregation algorithms exhibit somewhat better forecasting performance.

Predicting protein function using multiple kernels

High-throughput experimental techniques provide a wide variety of heterogeneous proteomic data sources. To exploit the information spread across multiple sources for protein function prediction, these data sources are transformed into kernels and then integrated into a composite kernel. Several methods first optimize the weights on these kernels to produce a composite kernel, and then train a classifier on the composite kernel. As such, these approaches result in an optimal composite kernel, but not necessarily in an optimal classifier. On the other hand, some approaches optimize the loss of binary classifiers and learn weights for the different kernels iteratively. For multi-class or multi-label data, these methods have to solve the problem of optimizing weights on these kernels for each of the labels, which are computationally expensive and ignore the correlation among labels. In this paper, we propose a method called Predicting Protein Function using Multiple K ernels (ProMK). ProMK iteratively optimizes the phases of learning optimal weights and reduces the empirical loss of multi-label classifier for each of the labels simultaneously. ProMK can integrate kernels selectively and downgrade the weights on noisy kernels. We investigate the performance of ProMK on several publicly available protein function prediction benchmarks and synthetic datasets. We show that the proposed approach performs better than previously proposed protein function prediction approaches that integrate multiple data sources and multi-label multiple kernel learning methods. The codes of our proposed method are available at https://sites.google.com/site/guoxian85/promk.

Asymmetric risk and return: evidence from the Australian Stock Exchange

This paper examines volatility asymmetry in a financial market using a stochastic volatility framework. We use the MCMC method for model estimations. There is evidence of volatility asymmetry in the data. Our asymmetric stochastic volatility in mean model, which nests both asymmetric stochastic volatility (ASV) and stochastic volatility in mean models (SVM), indicates ASV sufficiently captures the risk-return relationship; therefore, augmenting it with volatility in mean does not improve its performance. ASV fits the data better and yields more accurate out-of-sample forecasts than alternatives. We also demonstrate that asymmetry mainly emanates from the systematic parts of returns. As a result, it is more pronounced at the market level and the volatility feedback effect dominates the leverage effect.

Classification of team sport activities using a single wearable tracking device

Wearable tracking devices incorporating accelerometers and gyroscopes are increasingly being used for activity analysis in sports. However, minimal research exists relating to their ability to classify common activities. The purpose of this study was to determine whether data obtained from a single wearable tracking device can be used to classify team sport-related activities. Seventy-six non-elite sporting participants were tested during a simulated team sport circuit (involving stationary, walking, jogging, running, changing direction, counter-movement jumping, jumping for distance and tackling activities) in a laboratory setting. A MinimaxX S4 wearable tracking device was worn below the neck, in-line and dorsal to the first to fifth thoracic vertebrae of the spine, with tri-axial accelerometer and gyroscope data collected at 100Hz. Multiple time domain, frequency domain and custom features were extracted from each sensor using 0.5, 1.0, and 1.5s movement capture durations. Features were further screened using a combination of ANOVA and Lasso methods. Relevant features were used to classify the eight activities performed using the Random Forest (RF), Support Vector Machine (SVM) and Logistic Model Tree (LMT) algorithms. The LMT (79-92% classification accuracy) outperformed RF (32-43%) and SVM algorithms (27-40%), obtaining strongest performance using the full model (accelerometer and gyroscope inputs). Processing time can be reduced through feature selection methods (range 1.5-30.2%), however a trade-off exists between classification accuracy and processing time. Movement capture duration also had little impact on classification accuracy or processing time. In sporting scenarios where wearable tracking devices are employed, it is both possible and feasible to accurately classify team sport-related activities.

Stable feature selection with support vector machines

The support vector machine (SVM) is a popular method for classification, well known for finding the maximum-margin hyperplane. Combining SVM with l1-norm penalty further enables it to simultaneously perform feature selection and margin maximization within a single framework. However, l1-norm SVM shows instability in selecting features in presence of correlated features. We propose a new method to increase the stability of l1-norm SVM by encouraging similarities between feature weights based on feature correlations, which is captured via a feature covariance matrix. Our proposed method can capture both positive and negative correlations between features. We formulate the model as a convex optimization problem and propose a solution based on alternating minimization. Using both synthetic and real-world datasets, we show that our model achieves better stability and classification accuracy compared to several state-of-the-art regularized classification methods.

Robust network traffic classification

As a fundamental tool for network management and security, traffic classification has attracted increasing attention in recent years. A significant challenge to the robustness of classification performance comes from zero-day applications previously unknown in traffic classification systems. In this paper, we propose a new scheme of Robust statistical Traffic Classification (RTC) by combining supervised and unsupervised machine learning techniques to meet this challenge. The proposed RTC scheme has the capability of identifying the traffic of zero-day applications as well as accurately discriminating predefined application classes. In addition, we develop a new method for automating the RTC scheme parameters optimization process. The empirical study on real-world traffic data confirms the effectiveness of the proposed scheme. When zero-day applications are present, the classification performance of the new scheme is significantly better than four state-of-the-art methods: random forest, correlation-based classification, semi-supervised clustering, and one-class SVM.