Abstract: Early detection of anomalies in data centers is important to reduce downtimes and the costs of periodic maintenance. However, there is little research on this topic and even fewer on the fusion of sensor data for the detection of abnormal events. The goal of this paper is to propose a method for anomaly detection in data centers by combining sensor data (temperature, humidity, power) and deep learning models. The model described in the paper uses one autoencoder per sensor to reconstruct the inputs. The auto-encoders contain Long-Short Term Memory (LSTM) layers and are trained using the normal samples of the relevant sensors selected by correlation analysis. The difference signal between the input and its reconstruction is then used to classify the samples using feature extraction and a random forest classifier. The data measured by the sensors of a data center between January 2019 and May 2020 are used to train the model, while the data between June 2020 and May 2021 are used to assess it. Performances of the model are assessed a posteriori through F1-score by comparing detected anomalies with the data center’s history. The proposed model outperforms the state-of-the-art reconstruction method, which uses only one autoencoder taking multivariate sequences and detects an anomaly with a threshold on the reconstruction error, with an F1-score of 83.60% compared to 24.16%.
Abstract: Detecting subjectively biased statements is a vital task. This is because this kind of bias, when present in the text or other forms of information dissemination media such as news, social media, scientific texts, and encyclopedias, can weaken trust in the information and stir conflicts amongst consumers. Subjective bias detection is also critical for many Natural Language Processing (NLP) tasks like sentiment analysis, opinion identification, and bias neutralization. Having a system that can adequately detect subjectivity in text will boost research in the above-mentioned areas significantly. It can also come in handy for platforms like Wikipedia, where the use of neutral language is of importance. The goal of this work is to identify the subjectively biased language in text on a sentence level. With machine learning, we can solve complex AI problems, making it a good fit for the problem of subjective bias detection. A key step in this approach is to train a classifier based on BERT (Bidirectional Encoder Representations from Transformers) as upstream model. BERT by itself can be used as a classifier; however, in this study, we use BERT as data preprocessor as well as an embedding generator for a Bi-LSTM (Bidirectional Long Short-Term Memory) network incorporated with attention mechanism. This approach produces a deeper and better classifier. We evaluate the effectiveness of our model using the Wiki Neutrality Corpus (WNC), which was compiled from Wikipedia edits that removed various biased instances from sentences as a benchmark dataset, with which we also compare our model to existing approaches. Experimental analysis indicates an improved performance, as our model achieved state-of-the-art accuracy in detecting subjective bias. This study focuses on the English language, but the model can be fine-tuned to accommodate other languages.
Abstract: Uncontrolled growth of abnormal cells in the lung in the form of tumor can be either benign (non-cancerous) or malignant (cancerous). Patients with Lung Cancer (LC) have an average of five years life span expectancy provided diagnosis, detection and prediction, which reduces many treatment options to risk of invasive surgery increasing survival rate. Computed Tomography (CT), Positron Emission Tomography (PET), and Magnetic Resonance Imaging (MRI) for earlier detection of cancer are common. Gaussian filter along with median filter used for smoothing and noise removal, Histogram Equalization (HE) for image enhancement gives the best results without inviting further opinions. Lung cavities are extracted and the background portion other than two lung cavities is completely removed with right and left lungs segmented separately. Region properties measurements area, perimeter, diameter, centroid and eccentricity measured for the tumor segmented image, while texture is characterized by Gray-Level Co-occurrence Matrix (GLCM) functions, feature extraction provides Region of Interest (ROI) given as input to classifier. Two levels of classifications, K-Nearest Neighbor (KNN) is used for determining patient condition as normal or abnormal, while Artificial Neural Networks (ANN) is used for identifying the cancer stage is employed. Discrete Wavelet Transform (DWT) algorithm is used for the main feature extraction leading to best efficiency. The developed technology finds encouraging results for real time information and on line detection for future research.
Abstract: This paper aims at bringing a scientific contribution to the cardiac arrhythmia biomedical diagnosis systems; more precisely to the study of the amelioration of cardiac arrhythmia classification performance using artificial neural network, adaptive neuro-fuzzy and fuzzy inference systems classifiers. The purpose of this amelioration is to enable cardiologists to make reliable diagnosis through automatic cardiac arrhythmia analyzes and classifications based on high confidence classifiers. In this study, six classes of the most commonly encountered arrhythmias are considered: the Right Bundle Branch Block, the Left Bundle Branch Block, the Ventricular Extrasystole, the Auricular Extrasystole, the Atrial Fibrillation and the Normal Cardiac rate beat. From the electrocardiogram (ECG) extracted parameters, we constructed a matrix (360x360) serving as an input data sample for the classifiers based on neural networks and a matrix (1x6) for the classifier based on fuzzy logic. By varying three parameters (the quality of the neural network learning, the data size and the quality of the input parameters) the automatic classification permitted us to obtain the following performances: in terms of correct classification rate, 83.6% was obtained using the fuzzy logic based classifier, 99.7% using the neural network based classifier and 99.8% for the adaptive neuro-fuzzy based classifier. These results are based on signals containing at least 360 cardiac cycles. Based on the comparative analysis of the aforementioned three arrhythmia classifiers, the classifiers based on neural networks exhibit a better performance.
Abstract: The wireless communication network is developing
rapidly, thus the wireless security becomes more and more important.
Specific emitter identification (SEI) is an vital part of wireless
communication security as a technique to identify the unique
transmitters. In this paper, a SEI method based on multiscale
dispersion entropy (MDE) and refined composite multiscale dispersion
entropy (RCMDE) is proposed. The algorithms of MDE and RCMDE
are used to extract features for identification of five wireless
devices and cross-validation support vector machine (CV-SVM)
is used as the classifier. The experimental results show that the
total identification accuracy is 99.3%, even at low signal-to-noise
ratio(SNR) of 5dB, which proves that MDE and RCMDE can
describe the communication signal series well. In addition, compared
with other methods, the proposed method is effective and provides
better accuracy and stability for SEI.
Abstract: Early diagnosis of dental caries is essential for maintaining dental health. In this paper, method for diagnosis of dental caries is proposed using Laplacian filter, adaptive thresholding, texture analysis and Support Vector Machine (SVM) classifier. Analysis of the proposed method is compared with Otsu thresholding, watershed segmentation and active contouring method. Adaptive thresholding has comparatively better performance with 96.9% accuracy and 96.1% precision. The results are validated using statistical method, two-way ANOVA, at significant level of 5%, that shows the interaction of proposed method on performance parameter measures are significant. Hence the proposed technique could be used for detection of dental caries in automated computer assisted diagnosis system.
Abstract: Many research projects require accurate delineation of the different land cover type of the agricultural area. Especially it is critically important for the definition of specific plants like cannabis. However, the complexity of vegetation stands structure, abundant vegetation species, and the smooth transition between different seconder section stages make vegetation classification difficult when using traditional approaches such as the maximum likelihood classifier. Most of the time, classification distinguishes only between trees/annual or grain. It has been difficult to accurately determine the cannabis mixed with other plants. In this paper, a mixed distribution models approach is applied to classify pure and mix cannabis parcels using Worldview-2 imagery in the Lakes region of Turkey. Five different land use types (i.e. sunflower, maize, bare soil, and cannabis) were identified in the image. A constrained Gaussian mixture discriminant analysis (GMDA) was used to unmix the image. In the study, 255 reflectance ratios derived from spectral signatures of seven bands (Blue-Green-Yellow-Red-Rededge-NIR1-NIR2) were randomly arranged as 80% for training and 20% for test data. Gaussian mixed distribution model approach is proved to be an effective and convenient way to combine very high spatial resolution imagery for distinguishing cannabis vegetation. Based on the overall accuracies of the classification, the Gaussian mixed distribution model was found to be very successful to achieve image classification tasks. This approach is sensitive to capture the illegal cannabis planting areas in the large plain. This approach can also be used for monitoring and determination with spectral reflections in illegal cannabis planting areas.
Abstract: Classification of high resolution polarimetric Synthetic Aperture Radar (PolSAR) images plays an important role in land cover and land use management. Recently, classification algorithms based on Bag of Visual Words (BOVW) model have attracted significant interest among scholars and researchers in and out of the field of remote sensing. In this paper, BOVW model with pixel based low-level features has been implemented to classify a subset of San Francisco bay PolSAR image, acquired by RADARSAR 2 in C-band. We have used segment-based decision-making strategy and compared the result with the result of traditional Support Vector Machine (SVM) classifier. 90.95% overall accuracy of the classification with the proposed algorithm has shown that the proposed algorithm is comparable with the state-of-the-art methods. In addition to increase in the classification accuracy, the proposed method has decreased undesirable speckle effect of SAR images.
Abstract: Landcover classification using automated classification techniques, while employing remotely sensed multi-spectral imagery, is one of the promising areas of research. Different land conditions at different time are captured through satellite and monitored by applying different classification algorithms in specific environment. In this paper, a SPOT-5 image provided by SUPARCO has been studied and classified in Environment for Visual Interpretation (ENVI), a tool widely used in remote sensing. Then, Artificial Neural Network (ANN) classification technique is used to detect the land cover changes in Abbottabad district. Obtained results are compared with a pixel based Distance classifier. The results show that ANN gives the better overall accuracy of 99.20% and Kappa coefficient value of 0.98 over the Mahalanobis Distance Classifier.
Abstract: Electromyography (EMG) is one of the most important interfaces between humans and robots for rehabilitation. Decoding this signal helps to recognize muscle activation and converts it into smooth motion for the robots. Detecting each muscle’s pattern during walking and running is vital for improving the quality of a patient’s life. In this study, EMG data from 10 muscles in 10 subjects at 4 different speeds were analyzed. EMG signals are nonlinear with high dimensionality. To deal with this challenge, we extracted some features in time-frequency domain and used manifold learning and Laplacian Eigenmaps algorithm to find the intrinsic features that represent data in low-dimensional space. We then used the Bayesian classifier to identify various patterns of EMG signals for different muscles across a range of running speeds. The best result for vastus medialis muscle corresponds to 97.87±0.69 for sensitivity and 88.37±0.79 for specificity with 97.07±0.29 accuracy using Bayesian classifier. The results of this study provide important insight into human movement and its application for robotics research.
Abstract: Complementary and Alternative Medicine (CAM) techniques are quite popular and effective for chronic diseases. Iridology is more than 150 years old CAM technique which analyzes the patterns, tissue weakness, color, shape, structure, etc. for disease diagnosis. The objective of this paper is to validate the use of iridology for the diagnosis of the diabetes. The suggested model was applied in a systemic disease with ocular effects. 200 subject data of 100 each diabetic and non-diabetic were evaluated. Complete procedure was kept very simple and free from the involvement of any iridologist. From the normalized iris, the region of interest was cropped. All 63 features were extracted using statistical, texture analysis, and two-dimensional discrete wavelet transformation. A comparison of accuracies of six different classifiers has been presented. The result shows 89.66% accuracy by the random forest classifier.
Abstract: Abstract—Attribute or feature selection is one of the basic
strategies to improve the performances of data classification tasks,
and, at the same time, to reduce the complexity of classifiers,
and it is a particularly fundamental one when the number
of attributes is relatively high. Its application to unsupervised
classification is restricted to a limited number of experiments in
the literature. Evolutionary computation has already proven itself
to be a very effective choice to consistently reduce the number
of attributes towards a better classification rate and a simpler
semantic interpretation of the inferred classifiers. We present a feature
selection wrapper model composed by a multi-objective evolutionary
algorithm, the clustering method Expectation-Maximization (EM),
and the classifier C4.5 for the unsupervised classification of data
extracted from a psychological test named BASC-II (Behavior
Assessment System for Children - II ed.) with two objectives:
Maximizing the likelihood of the clustering model and maximizing
the accuracy of the obtained classifier. We present a methodology
to integrate feature selection for unsupervised classification, model
evaluation, decision making (to choose the most satisfactory model
according to a a posteriori process in a multi-objective context), and
testing. We compare the performance of the classifier obtained by the
multi-objective evolutionary algorithms ENORA and NSGA-II, and
the best solution is then validated by the psychologists that collected
the data.
Abstract: In recent years, there has been an explosion in the rate of using technology that help discovering the diseases. For example, DNA microarrays allow us for the first time to obtain a "global" view of the cell. It has great potential to provide accurate medical diagnosis, to help in finding the right treatment and cure for many diseases. Various classification algorithms can be applied on such micro-array datasets to devise methods that can predict the occurrence of Leukemia disease. In this study, we compared the classification accuracy and response time among eleven decision tree methods and six rule classifier methods using five performance criteria. The experiment results show that the performance of Random Tree is producing better result. Also it takes lowest time to build model in tree classifier. The classification rules algorithms such as nearest- neighbor-like algorithm (NNge) is the best algorithm due to the high accuracy and it takes lowest time to build model in classification.
Abstract: In this paper, we present a pedestrian detection descriptor called Fused Structure and Texture (FST) features based on the combination of the local phase information with the texture features. Since the phase of the signal conveys more structural information than the magnitude, the phase congruency concept is used to capture the structural features. On the other hand, the Center-Symmetric Local Binary Pattern (CSLBP) approach is used to capture the texture information of the image. The dimension less quantity of the phase congruency and the robustness of the CSLBP operator on the flat images, as well as the blur and illumination changes, lead the proposed descriptor to be more robust and less sensitive to the light variations. The proposed descriptor can be formed by extracting the phase congruency and the CSLBP values of each pixel of the image with respect to its neighborhood. The histogram of the oriented phase and the histogram of the CSLBP values for the local regions in the image are computed and concatenated to construct the FST descriptor. Several experiments were conducted on INRIA and the low resolution DaimlerChrysler datasets to evaluate the detection performance of the pedestrian detection system that is based on the FST descriptor. A linear Support Vector Machine (SVM) is used to train the pedestrian classifier. These experiments showed that the proposed FST descriptor has better detection performance over a set of state of the art feature extraction methodologies.
Abstract: We assume an IoT-based smart-home environment where the on-off status of each of the electrical appliances including the room lights can be recognized in a real time by monitoring and analyzing the smart meter data. At any moment in such an environment, we can recognize what the household or the user is doing by referring to the status data of the appliances. In this paper, we focus on a smart-home service that is to activate a robot vacuum cleaner at right time by recognizing the user situation, which requires a situation-aware model that can distinguish the situations that allow vacuum cleaning (Yes) from those that do not (No). We learn as our candidate models a few classifiers such as naïve Bayes, decision tree, and logistic regression that can map the appliance-status data into Yes and No situations. Our training and test data are obtained from simulations of user behaviors, in which a sequence of user situations such as cooking, eating, dish washing, and so on is generated with the status of the relevant appliances changed in accordance with the situation changes. During the simulation, both the situation transition and the resulting appliance status are determined stochastically. To compare the performances of the aforementioned classifiers we obtain their learning curves for different types of users through simulations. The result of our empirical study reveals that naïve Bayes achieves a slightly better classification accuracy than the other compared classifiers.
Abstract: Health diseases have a vital significance affecting human being's life and life quality. Sudden death events can be prevented owing to early diagnosis and treatment methods. Electrical signals, taken from the human being's body using non-invasive methods and showing the heart activity is called Electrocardiogram (ECG). The ECG signal is used for following daily activity of the heart by clinicians. Heart Rate Variability (HRV) is a physiological parameter giving the variation between the heart beats. ECG data taken from MITBIH Arrhythmia Database is used in the model employed in this study. The detection of arrhythmic heart beats is aimed utilizing the features extracted from the HRV time domain parameters. The developed model provides a satisfactory performance with ~89% accuracy, 91.7 % sensitivity and 85% specificity rates for the detection of arrhythmic beats.
Abstract: One of the most critical decision points in the design of a
face recognition system is the choice of an appropriate face representation.
Effective feature descriptors are expected to convey sufficient, invariant
and non-redundant facial information. In this work we propose a set of
Hahn moments as a new approach for feature description. Hahn moments
have been widely used in image analysis due to their invariance, nonredundancy
and the ability to extract features either globally and locally.
To assess the applicability of Hahn moments to Face Recognition we
conduct two experiments on the Olivetti Research Laboratory (ORL)
database and University of Notre-Dame (UND) X1 biometric collection.
Fusion of the global features along with the features from local facial
regions are used as an input for the conventional k-NN classifier. The
method reaches an accuracy of 93% of correctly recognized subjects for
the ORL database and 94% for the UND database.
Abstract: Margin-Based Principle has been proposed for a long
time, it has been proved that this principle could reduce the
structural risk and improve the performance in both theoretical
and practical aspects. Meanwhile, feed-forward neural network is
a traditional classifier, which is very hot at present with a deeper
architecture. However, the training algorithm of feed-forward neural
network is developed and generated from Widrow-Hoff Principle that
means to minimize the squared error. In this paper, we propose
a new training algorithm for feed-forward neural networks based
on Margin-Based Principle, which could effectively promote the
accuracy and generalization ability of neural network classifiers
with less labelled samples and flexible network. We have conducted
experiments on four UCI open datasets and achieved good results
as expected. In conclusion, our model could handle more sparse
labelled and more high-dimension dataset in a high accuracy while
modification from old ANN method to our method is easy and almost
free of work.
Abstract: The performance and analysis of speech recognition
system is illustrated in this paper. An approach to recognize the
English word corresponding to digit (0-9) spoken by 2 different
speakers is captured in noise free environment. For feature extraction,
speech Mel frequency cepstral coefficients (MFCC) has been used
which gives a set of feature vectors from recorded speech samples.
Neural network model is used to enhance the recognition
performance. Feed forward neural network with back propagation
algorithm model is used. However other speech recognition
techniques such as HMM, DTW exist. All experiments are carried
out on Matlab.
Abstract: The goal of image segmentation is to cluster pixels
into salient image regions. Segmentation could be used for object
recognition, occlusion boundary estimation within motion or stereo
systems, image compression, image editing, or image database lookup.
In this paper, we present a color image segmentation using
support vector machine (SVM) pixel classification. Firstly, the pixel
level color and texture features of the image are extracted and they
are used as input to the SVM classifier. These features are extracted
using the homogeneity model and Gabor Filter. With the extracted
pixel level features, the SVM Classifier is trained by using FCM
(Fuzzy C-Means).The image segmentation takes the advantage of
both the pixel level information of the image and also the ability of
the SVM Classifier. The Experiments show that the proposed method
has a very good segmentation result and a better efficiency, increases
the quality of the image segmentation compared with the other
segmentation methods proposed in the literature.