Two-class Classification Problem Research Articles

Algoritma Multi-Kelas Twin Bounded SVM Untuk Klasifikasi Pola

Pattern recognition is a process of recognizing patterns by using machine learning algorithm. Pattern recognition can be defined as a classification of data based on knowledge that already gained or information extracted from patterns. One method that can be used in pattern classification problem is SVM. In this study we introduced Twin Bounded SVM which is refinement of Twin SVM. The discussion begins with the linear Twin Bounded SVM method to solve a two-class classification problem and followed by an algorithm to solve multi-class classification problem

Fingerprint liveness detection based on guided filtering and hybrid image analysis

Fingerprints are widely used for biometric recognition. However, many spoofing attacks based on an artificially made fingerprint occur. In this study, the authors propose an approach to detect fingerprint liveness which uses the guided filtering and hybrid image analysis. This study deals with the problem of ignoring the contribution that is brought by the sharp features when analysing the denoised image. The method described utilises both the enhanced sharp features and denoised features from the hybrid images to get better results. The input fingerprint is pre-processed by region of interest extraction and then is filtered by a guidance image for obtaining the denoised image. Then, histogram equalisation is introduced to eliminate the impact of illumination condition. The authors extract the co-occurrence of adjacent local binary pattern features from both the cropped images and the denoised images. Whilst concatenating both the features together to form a long feature, t-Distributed Stochastic Neighbour Embedding is applied to reduce the data dimension. The authors consider the fingerprint liveness detection as a two-class classification problem and use support vector machine with radial basis function kernel to solve this problem. The authors evaluate the experiments on three benchmark data sets. Experimental results demonstrate that the accuracy of the proposed method can outperform most of the state-of-art methods.

Spam review detection using LSTM autoencoder: an unsupervised approach

The review of online products or services is becoming a major factor in the user’s purchasing decisions. The popularity and influence of online reviews attract spammers who intend to elevate their products or services by writing positive reviews for them and lowering the business of others by writing negative reviews. Traditionally, the spam review identification task is seen as a two-class classification problem. The classification approach requires a labelled dataset to train a model for the environment it is working on. The unavailability of the labelled dataset is a major limitation in the classification approach. To overcome the problem of the labelled dataset, we propose an unsupervised learning model combining long short-term memory (LSTM) networks and autoencoder (LSTM-autoencoder) to distinguish spam reviews from other real reviews. The said model is trained to learn the patterns of real review from the review’s textual details without any label. The experimental results show that our model is able to separate the real and spam review with good accuracy.

A Novel Deep Neural Network for Robust Detection of Seizures Using EEG Signals

The detection of recorded epileptic seizure activity in electroencephalogram (EEG) segments is crucial for the classification of seizures. Manual recognition is a time-consuming and laborious process that places a heavy burden on neurologists, and hence, the automatic identification of epilepsy has become an important issue. Traditional EEG recognition models largely depend on artificial experience and are of weak generalization ability. To break these limitations, we propose a novel one-dimensional deep neural network for robust detection of seizures, which composes of three convolutional blocks and three fully connected layers. Thereinto, each convolutional block consists of five types of layers: convolutional layer, batch normalization layer, nonlinear activation layer, dropout layer, and max-pooling layer. Model performance is evaluated on the University of Bonn dataset, which achieves the accuracy of 97.63%∼99.52% in the two-class classification problem, 96.73%∼98.06% in the three-class EEG classification problem, and 93.55% in classifying the complicated five-class problem.

Optical classifier based on volume holographic filter with vector decomposition

A new implementation of the weight vector in a classifier based on an optical system using a volume holographic filter with a vector decomposition algorithm is proposed for high-speed data classification. Using a volume holographic filter, the correlation between the input and the decomposed weight vector is calculated for the classification score, which is the output from the decision function in the optical classifier. The authors experimentally demonstrate the core concept and confirm the effectiveness of the proposed method by executing a two-class classification problem using a volume hologram medium.

Change detection using least squares one-class classification control chart

ABSTRACT One-class classification can be thought as a special type of two-class classification problem, where data only from one class, the target class, are available for training the classifier (referred to as one-class classifier). The problem of classifying positive (or target) cases in the absence of appropriately characterized negative cases (or outliers) has gained increasing attention in recent years. Several methods are available to solve the one-class classification problem. Three methods are commonly used: density estimation, boundary methods, and reconstruction methods. This paper focuses on boundary methods which include k–center method, nearest neighbor method, one-class support vector machine (OCSVM), and support vector data description (SVDD). In statistical process control (SPC), practitioners successfully used SVDD to detect anomalies or outliers in the process. In this paper, we reformulate the standard OCSVM by a least squares version of the method. This least squares one-class support vector machine (LS-OCSVM) is used to design a control chart for monitoring the mean vector of processes. We compare the performance of the LS-OCSVM chart with the SVDD and chart. The experimental results indicate that the proposed control chart has very good performances.

Tree-Based Contrast Subspace Mining for Categorical Data

Mining contrast subspace has emerged to find subspaces where a particular queried object is most similar to the target class against the non-target class in a two-class data set. It is important to discover those subspaces, which are known as contrast subspaces, in many real-life applications. Tree-Based Contrast Subspace Miner (TB-CSMiner) method has been recently introduced to mine contrast subspaces of queried objects specifically for numerical data set. This method employs tree-based scoring function to estimate the likelihood contrast score of subspaces with respect to the given queried object. However, it limits the use of TB-CSMiner on categorical values that are frequently encountered in real-world data sets. In this paper, the TB-CSMiner method is extended by formulating the tree-based likelihood contrast scoring function for mining contrast subspace in categorical data set. The extended method uses features values of queried object to gather target samples having similar characteristics into the same group and separate non-target samples having different characteristics from this queried object in different group. Given a contrast subspace of the target samples, the queried object should fall in a group having target samples more than the non-target samples. Several experiments have been conducted on eight real world categorical data sets to evaluate the effectiveness of the proposed extended TB-CSMiner method by performing classification tasks in a two-class classification problem with categorical input variables. The obtained results demonstrated that the extended method can improve the performance accuracy of most classification tasks. Thus, the proposed extended tree-based method is also shown to have the ability to discover contrast subspaces of the given queried object in categorical data.

Providing a new method for link prediction in social networks based on the meta-heuristic algorithm

Social network analysis is one of the most important research fields in data mining today. The purpose of the analysis of these networks is to extract the embedded knowledge in the data set and to learn the behavior of users in the social networking environment. One of the most attractive and central applications of social network analysis is link prediction. The purpose of link prediction in social networks is to identify missing and unknown information from users or to predict the future link between two users. In recent years, various artificial intelligence algorithms have been introduced as one of the most important tools for resolving link prediction and big data. In this research, a strategy based on a meta-heuristic is used to improve link prediction in social networks. The proposed method is based on the characteristics of the signed social networks provided and turns the link prediction problem into a two-class classification problem. Then uses the capability of the Particle Swarm Optimization (PSO) and the topological properties of the social network graph to create a database with two classes, the first class pointing to the existence of a connection between the users and the second class pointing to the absence of this relationship. Creates a database using the support vector machine model for categorization work and uses the classic Katz similarity criterion for end-user suggestions. Twitter social network information has been used to compare and evaluate the proposed method. The results of the experiments show the superiority of the proposed method with 0.23, 0.99, and 6.32, respectively, compared to Meta-Path, Katz and CN algorithms in F-measure criterion.

SUPERVISED DETECTION OF BOMB CRATERS IN HISTORICAL AERIAL IMAGES USING CONVOLUTIONAL NEURAL NETWORKS

Abstract. The aftermath of the air strikes during World War II is still present today. Numerous bombs dropped by planes did not explode, still exist in the ground and pose a considerable explosion hazard. Tracking down these duds can be tackled by detecting bomb craters. The existence of a dud can be inferred from the existence of a crater. This work proposes a method for the automatic detection of bomb craters in aerial wartime images. First of all, crater candidates are extracted from an image using a blob detector. Based on given crater references, for every candidate it is checked whether it, in fact, represents a crater or not. Candidates from various aerial images are used to train, validate and test Convolutional Neural Networks (CNNs) in the context of a two-class classification problem. A loss function (controlling what the CNNs are learning) is adapted to the given task. The trained CNNs are then used for the classification of crater candidates. Our work focuses on the classification of crater candidates and we investigate if combining data from related domains is beneficial for the classification. We achieve a F1-score of up to 65.4% when classifying crater candidates with a realistic class distribution.

Clustering-based undersampling with random over sampling examples and support vector machine for imbalanced classification of breast cancer diagnosis

To overcome the two-class imbalanced classification problem existing in the diagnosis of breast cancer, a hybrid of Random Over Sampling Example, K-means and Support vector machine (RK-SVM) model is proposed which is based on sample selection. Random Over Sampling Example (ROSE) is utilized to balance the dataset and further improve the diagnosis accuracy by Support Vector Machine (SVM). As there is one different sample selection factor via clustering that encourages selecting the samples near the class boundary. The purpose of clustering here is to reduce the risk of removing useful samples and improve the efficiency of sample selection. To test the performance of the new hybrid classifier, it is implemented on breast cancer datasets and the other three datasets from the University of California Irvine (UCI) machine learning repository, which are commonly used datasets in class imbalanced learning. The extensive experimental results show that our proposed hybrid method outperforms most of the competitive algorithms in term of G-mean and accuracy indices. Additionally, experimental results show that this method also performs superiorly for binary problems.

Hardware Trojan Detection Leveraging a Novel Golden Layout Model Towards Practical Applications

Globalization trend in integrated circuit design and manufacturing process has increased the vulnerability of integrated circuit. These vulnerabilities mainly caused by hardware Trojan have a serious impact on the security of integrated circuits. Although side-channel analysis approach is the most promising Trojan detection approach, nearly all side-channel analysis approaches rely heavily on the availability of golden chips, which are extremely difficult to obtain. In this paper, a golden layout model instead of fabricated golden chips is introduced for the practical application of hardware Trojan detection approaches. The simulated voltage variations generated from the golden layout model at different process corners serve as golden reference, thus fabricated golden chips are not required during detection. Further, silicon measurements are performed to obtain the voltage variations of fabricated chips, and a model calibration algorithm is utilized to calibrate the golden model in the presence of process variations and random noise. Finally, the Trojan detection is formulated as a two-class classification problem, and the Trojan is identified using the partitioning around medoids algorithm. Experimental results demonstrate that the similarities between the simulated traces and measured traces are greater than 98.81%, and the proposed approach distinguishes the Trojan chips correctly even under ± 15% process variation.

Real-Time Detection of Fall From Bed Using a Single Depth Camera

Toward the medical and living healthcare for the elderly and patients, fall from bed is a critical accident that may lead to serious injuries. To alleviate this, an essential problem is to detect this event in time for earning the rescue time. Although some efforts that resort to the wearable devices and smart healthcare room have already been paid to address this problem, the performance is still not satisfactory enough for the practical applications. In this paper, a novel fall from a bed detection method is proposed. In particular, the depth camera is used as the visual sensor due to its insensitivity to illumination variation and capacity of privacy protection. To characterize the human activity well, an effective human upper body detection approach able to extract human head and upper body center is proposed using random forest. Compared with the existing widely used human body parsing methods (e.g., Microsoft Kinect SDK or OpenNI SDK), our proposition can still work reliably when human–bed interaction happens. According to the motion information of human upper body, the fall from bed detection task is formulated as a two-class classification problem. Then, it is solved using the large margin nearest neighbor classification approach. Our method can meet the real-time running requirement with the normal computer. In experiments, we construct a fall from bed detection data set that contains the samples from 42 volunteers (26 males and 16 females) for test. The experimental results demonstrate the effectiveness and efficiency of our proposition. Note to Practitioners—This paper aims to develop the vision-based surveillance system to automatically detect fall from bed in real time under the unconstrained conditions. This cannot be well handled by using wearable devices. Our proposition can be used to construct a smart healthcare room. It is worth noting that the proposed system is insensitive to illumination variation and capable of protecting privacy due to the use of depth camera. The proposed human body extraction method is able to address the human–bed interaction well. An effective motion pattern categorization approach is applied to ensure the good separation between fall from bed and the other activities. According to the extensive experiments under the laboratory and sickroom environments, our fall from bed detection method generally can achieve acceptable result with high efficiency. Nevertheless, it is still somewhat sensitive to noise within the depth frames. In addition, when serious human–quilt interaction happens, the performance of human body extraction procedure is not satisfactory enough.

Single-Sentence Compression using XGBoost

Sentence compression is known as presenting a sentence in a fewer number of words compared to its original one without changing the meaning. Recent works on sentence compression formulates the problem as an integer linear programming problem (ILP) then solves it using an external ILP-solver which suffers from slow running time. In this article, the sentence compression task is formulated as a two-class classification problem and used a gradient boosting technique to solve the problem. Different models are created using two different datasets. The best model is taken for evaluation. The quality of compression is measured using two important quality measures, informativeness and compression rate. This article has achieved 70.2 percent in informativeness and 38.62 percent in compression rate.

Epileptic seizure identification using entropy of FBSE based EEG rhythms

This paper has proposed a new method for classification of epileptic seizures based on weighted multiscale Renyi permutation entropy (WMRPE) and rhythms obtained with Fourier–Bessel series expansion (FBSE) of electroencephalogram (EEG) signals. In this method, each EEG signal is considered for the FBSE to determine the coefficients then the selected order values of these coefficients are used to obtain the rhythms (δ, θ, α, β, and γ). The feature used in this work is WMRPE which has been extracted from rhythms of EEG signal. These computed feature values are then used in distinct classifiers such as random forest (RF), least squares support vector machine (LS-SVM), and regression for classification of epileptic seizure EEG signals. The classification is validated using 10-fold cross-validation technique in order to verify the robustness of the proposed work and the optimization of classification performance has been implemented with feature ranking methods. The proposed method is also tested with additive white Gaussian noise (AWGN) at different signal to noise ratio (SNR) levels. In this present work, the proposed method provides better classification accuracy results for seven different classification problems in which six are two-class classification problems and the remaining one is three-class classification problem. In two-class classification problems, the EEG signals of epileptic patients at the time of seizure interval has been classified with six distinct classes of EEG signals i.e., healthy subjects with eyes open condition, healthy subjects with eyes closed condition, epileptic patients at the time of seizure-free interval captured from hippocampal formation zone, epileptic patients at the time of seizure-free interval captured from epileptogenic zone, epileptic patients at the time of seizure-free interval, and subjects at the time of non-seizure (healthy and seizure-free) interval. In three-class classification problem, the EEG signals corresponding to epileptic patients during seizure interval, healthy subjects with eyes open condition, and epileptic patients at the time of seizure-free interval captured from hippocampal formation zone have been classified with the help of proposed method. The obtained results show that the WMRPE with FBSE based rhythms provides better classification accuracies in comparison to the existing rhythms based method and other existing methods. The proposed method is also perform well in noisy environment at different SNR levels.

Hierarchical method for cataract grading based on retinal images using improved Haar wavelet

Cataracts, which are lenticular opacities that may occur at different lens locations, are the leading cause of visual impairment worldwide. Accurate and timely diagnosis can improve the quality of life of cataract patients. In this paper, a feature extraction-based method for grading cataract severity using retinal images is proposed. To obtain more appropriate features for the automatic grading, the Haar wavelet is improved according to the characteristics of retinal images. Retinal images of non-cataract, as well as mild, moderate, and severe cataracts, are automatically recognized using the improved Haar wavelet. A hierarchical strategy is used to transform the four-class classification problem into three adjacent two-class classification problems. Three sets of two-class classifiers based on a neural network are trained individually and integrated together to establish a complete classification system. The accuracies of the two-class classification (cataract and non-cataract) and four-class classification are 94.83% and 85.98%, respectively. The performance analysis demonstrates that the improved Haar wavelet feature achieves higher accuracy than the original Haar wavelet feature, and the fusion of three sets of two-class classifiers is superior to a simple four-class classifier. The discussion indicates that the retinal image-based method offers significant potential for cataract detection.

SOLVING TWO-CLASS CLASSIFICATION PROBLEM USING MEMETIC PROGRAMMING

SOLVING TWO-CLASS CLASSIFICATION PROBLEM USING MEMETIC PROGRAMMING

Steganalysis aided by fragile detection of image manipulations

Steganalysis is usually considered as a two-class classification problem of differentiating between covers and stegos. However, in the real world, the cover image may have undergone various operations, which causes two problems that some processed covers tend to be judged as stegos by the steganalyzer and the stegos processed before information embedding may be easily missed, resulting in the high false alarm rate and the high missed detection rate of steganalysis respectively. To address the former problem, this paper proposed a steganalysis framework based on the combination of the image forensics and the steganalysis tools to reduce the false alarms. First, the fragile detection of image manipulations which is not robust to steganography is applied to separate the normally processed images from the investigated images. Then remaining images are fed to the trained classifier for stegnalysis. The experimental results on gamma transformed images validate the effectiveness of the proposed steganalysis framework that the false alarm rates of steganalysis can be reduced when the investigated image dataset contains normally processed images.

LMR: Learning a Two-Class Classifier for Mismatch Removal.

Feature matching, which refers to establishing reliable correspondence between two sets of features, is a critical prerequisite in a wide spectrum of vision-based tasks. Existing attempts typically involve the mismatch removal from a set of putative matches based on estimating the underlying image transformation. However, the transformation could vary with different data. Thus, a pre-defined transformation model is often demanded, which severely limits the applicability. From a novel perspective, this paper casts the mismatch removal into a two-class classification problem, learning a general classifier to determine the correctness of an arbitrary putative match, termed as Learning for Mismatch Removal (LMR). The classifier is trained based on a general match representation associated with each putative match through exploiting the consensus of local neighborhood structures based on a multiple K -nearest neighbors strategy. With only ten training image pairs involving about 8000 putative matches, the learned classifier can generate promising matching results in linearithmic time complexity on arbitrary testing data. The generality and robustness of our approach are verified under several representative supervised learning techniques as well as on different training and testing data. Extensive experiments on feature matching, visual homing, and near-duplicate image retrieval are conducted to reveal the superiority of our LMR over the state-of-the-art competitors.

An Effective Metaheuristic for Bi-objective Feature Selection in Two-Class Classification Problem

Feature selection is known as a very useful technique in machine learning practice as it may result in the development of more straightforward models with better accuracy. Traditionally, feature selection is considered as a single-objective problem, however, it can be easily formulated in terms of two objectives. The solving of such problems requires the application of appropriate multi-objective optimization methods that do not always offer equally good solutions even under the same conditions. This paper focuses on the development of a metaheuristic optimization approach for bi-objective feature selection problem in two-class classification. We consider the solving of this problem in terms of minimization of both misclassification error and feature subset size. For solving the considered problem, an adaptation of the Multi-Objective Adaptive Memory Programming (MOAMP) metaheuristic based on the tabu search strategy is proposed. Our MOAMP adaption has been utilized to obtain the sets of most relevant features for two real classification problems with two classes. Finally, using popular Pareto front quality indicators, the obtained results have been compared with the sets of non-dominated solutions derived by the well-known NSGA2 algorithm. The conducted research allows concluding about the ability of the MOAMP adaptation to get a better efficient frontier for the same number of objective function calls.

A Novel Neural Network for Remote Sensing Image Matching.

Rapid development of remote sensing (RS) imaging technology makes the acquired images have larger size, higher resolution, and more complex structure, which goes beyond the reach of classical hand-crafted feature-based matching. In this paper, we propose a feature learning approach based on two-branch networks to transform the image matching task into a two-class classification problem. To match two key points, two image patches centered at the key points are entered into the proposed network. The network aims to learn discriminative feature representations for patch matching, so that more matching pairs can be obtained on the premise of maintaining higher subpixel matching accuracy. The proposed network adopts a two-stage training mode to deal with the complex characteristics of RS images. An adaptive sample selection strategy is proposed to determine the size of each patch by the scale of its central key point. Thus, each patch can preserve the texture structure around its key point rather than all patches have a predetermined size. In the matching prediction stage, two strategies, namely, superpixel-based sample graded strategy and superpixel-based ordered spatial matching, are designed to improve the matching efficiency and matching accuracy, respectively. The experimental results and theoretical analysis demonstrate the feasibility, robustness, and effectiveness of the proposed method.