Search

In order to solve the problem of low accuracy of time dimension feature extraction and classification of high-dimensional large data streams, this paper proposes a time dimension feature extraction and classification algorithm of high-dimensional large data streams based on unsupervised learning. Analyze the trend of high-dimensional data flow changes under machine learning, and achieve dimensionality reduction of high-dimensional large traffic time dimensional data through local save projection. Analyze the spatial relationship between feature attributes and feature space, segment and fit high-dimensional big data streams and time dimensional feature data streams, further segment time dimensional sequences using sliding windows, and complete feature extraction through discrete dyadic wavelet transform. According to the clustering algorithm, cluster the time dimension feature data stream, calculate the cosine similarity of the feature data, model the time dimension feature stream of training samples, use the feature classification function to minimize the classification loss, and use unsupervised learning to achieve the final classification task. The test results show that this method can improve the temporal feature extraction and classification accuracy streams. [ABSTRACT FROM AUTHOR]

Adverse weather conditions, such as snow-covered roads, represent a challenge for autonomous vehicle research. This is particularly challenging as it might cause misalignment between the longitudinal axis of the vehicle and the actual direction of travel. In this paper, we extend previous work in the field of autonomous vehicles on snow-covered roads and present a novel approach for side-slip angle estimation that combines perception with a hybrid artificial neural network pushing the prediction horizon beyond existing approaches. We exploited the feature extraction capabilities of convolutional neural networks and the dynamic time series relationship learning capabilities of gated recurrent units and combined them with a motion model to estimate the side-slip angle. Subsequently, we evaluated the model using the 3DCoAutoSim simulation platform, where we designed a suitable simulation environment with snowfall, friction, and car tracks in snow. The results revealed that our approach outperforms the baseline model for prediction horizons ⩾ 2 seconds. This extended prediction horizon has practical implications, by providing drivers and autonomous systems with more time to make informed decisions, thereby enhancing road safety. [ABSTRACT FROM AUTHOR]

News text is an important branch of natural language processing. Compared to ordinary texts, news text has significant economic and scientific value. The characteristics of news text include structural hierarchy, diverse label categories, and limited high-quality annotation samples. Many machine learning and deep learning methods exist to analyze various forms of news text. However, due to label imbalance, hierarchical semantics, and confusing labels, current methods have limitations. Therefore, this paper proposes a news text classification framework based on hierarchical semantics and prior correction (HSPC). Firstly, data augmentation is used to enhance the diversity of the training set and adversarial learning is employed to improve the resistance of the model with its robustness. Then, a hierarchical feature extraction approach is employed to extract semantic features from different levels of news texts. Consequentially, a feature fusion method is designed to allow the model to focus on relevant hierarchical semantics for label classification. Finally, highly confusing label predictions are corrected to optimize the label prediction of the model and improve confidence. Multiple experiments are performed on four widely used public datasets. The experimental results indicate that HSPC achieves higher classification accuracy compared to other models. On the FCT, AGNews, THUCNews, and Ohsumed datasets, HSPC improves the accuracy by 1.03%, 1.38%, 2.55%, and 1.15%, respectively, compared to state-of-the-art methods. This validates the rationality and effectiveness of the designed mechanisms. [ABSTRACT FROM AUTHOR]

Neuroimaging technology is considered a non-invasive method research the structure and function of the brain which have been widely used in neuroscience, psychiatry, psychology, and other fields. The development of Deep Learning Neural Network (DLNN), based on the deep learning algorithms of neural imaging techniques in brain disease diagnosis plays a more and more important role. In this paper, a deep neural network imaging technology based on Stack Auto-Encoder (SAE) feature extraction is constructed, and then Support Vector Machine (SVM) was used to solve binary classification problems (Alzheimer's disease [AD] and Mild Cognitive Impairment [MCI]). Four sets of experimental data were employed to perform the training and testing stages of DLNN. The number of neurons in each of the DLNNs was determined using the grid search technique. Overall, the results of DLNNs performance indicated that the SAE feature extraction was superior over (Accuracy Rate [AR] = 74.9% with structure of 93-171-49-22-93) shallow layer features extraction (AR = 70.8% with structure of 93-22-93) and primary features extraction (AR = 69.2%). [ABSTRACT FROM AUTHOR]

In the field of text classification, current research ignores the role of part-of-speech features, and the multi-channel model that can learn richer text information compared to a single model. Moreover, the method based on neural network models to achieve final classification, using fully connected layer and Softmax layer can be further improved and optimized. This paper proposes a hybrid model for text classification using part-of-speech features, namely PAGNN-Stacking1. In the text representation stage of the model, introducing part-of-speech features facilitates a more accurate representation of text information. In the feature extraction stage of the model, using the multi-channel attention gated neural network model can fully learn the text information. In the text final classification stage of the model, this paper innovatively adopts Stacking algorithm to improve the fully connected layer and Softmax layer, which fuses five machine learning algorithms as base classifier and uses fully connected layer Softmax layer as meta classifier. The experiments on the IMDB, SST-2, and AG_News datasets show that the accuracy of the PAGNN-Stacking model is significantly improved compared to the benchmark models. [ABSTRACT FROM AUTHOR]

From the signal received on a particular frequency band, spectrum sensing (SS) is used in cognitive radio (CR) to assess whether the primary user (PU) is using the spectrum and, consequently, whether the secondary user (SU) can utilize the spectrum. The main issue with SS is determining the presence of the primary signal in a low signal-to-noise ratio (SNR). Compared to conventional technologies, machine learning techniques are more effective and accurate at identifying the qualities of input data. This paper proposes a machine learning (ML) based SS model for CR with effective feature extraction and reduction techniques. The proposed work comprises five phases: noise removal, wavelet transform, feature extraction, dimensionality reduction, and classification. Firstly, noise filtering is done on the received signal to remove the noise present in the input signal using the filters such as moving median filter (MMF), Gaussian filter (GF), and Gabor filter (GBF). After that, the filtered signal is transformed into a wavelet domain using Discrete Wavelet Transform (DWT) algorithm. Then the statistical features such as average absolute value, wavelet energy, variance, standard deviation, and peak value features are extracted from the DWT. Next, the dimensionality reduction (DR) is performed using Linear Discriminant Analysis (LDA). Finally, the classification is performed using the ensemble ML classifiers such as Support Vector Machine (SVM), Naive Bayes (NB), and K-Nearest Neighbour (KNN), which classify whether the PU signal is active or not. Simulations are carried out to analyze the efficiency of the presented models for SS. The results proved that SVM obtains the best performance for SS with higher accuracy and lower SNR. [ABSTRACT FROM AUTHOR]

Healthcare data is the most sensitive information for processing through machine learning and cloud computing in the various healthcare organizations. Electronic Health Record (EHR) manipulation are now on the rise, and we need to focus on using the data generated by the healthcare applications. Many sensitive data are associated with various health care domains, particularly neurology and cardiology. Previous approaches, such as manual data records, had significant disadvantages, and hence disease prediction based on the above records was found ineffective resulting with improper diagnosis on the patients. These data records require special attention, and current frameworks focused on these areas must implement sophisticated technologies to predict specific patterns. To address the above concerns, the proposed work incorporates the integration of Neuro Fuzzy Logistic Regression (NFLR) machine learning algorithm and cloud computing storage management to solve these problems. The usage of cloud storage reduces data duplication while handling the storage of EHRs where the proposed ML algorithm accurately predict the disease. In the proposed research, the features are extracted using a specific algorithm –Self-organizing Clustering (SOC) which forms a clustered data with highest weight. To select the maximum number of features, and to predict the disease risk factors, the S2NO algorithm and NFLR algorithms are used in this work. Further, the database storage estimation with fuzzy rules, logistic analysis, and other benefits such as experimental learning of different ML tools, data privacy constraints related to healthcare are considered in this paper. [ABSTRACT FROM AUTHOR]

This paper proposed a hybrid intelligent approach based on empirical mode decomposition (EMD), artificial neural network (ANN) and J48 algorithm of machine learning for real-time harmonics analysis of digital substation's equipment based on IEC-61850 using explanatory input variables based on laboratory proto-type real-time recorded database. In the proposed hybrid model, these variables are first extracted then diagnostic of power transformer harmonics of digital substation is evaluated/analyzed to perform the long term as well as the short term goal and planning in the electrical power network. In this paper, firstly, experimental analysis is performed to validate the laboratory prototype setup using FFT (fast Fourier transform), STFT (short-time Fourier transform) and CWT (continuous wavelet transform). Then, features are extracted from experimental dataset using EMD (empirical mode decomposition) method. The IMFs (intrinsic mode functions) have generated from EMD, which are used as an input variable to the two different diagnostic models, i.e., ANN and J48 algorithm. In order to validate the performance and accuracy of the proposed hybrid model, a comparative analysis is performed by using ANN and J48 method (with and without EMD method) and the results are compared. Obtained results shows that the proposed hybrid diagnostics approach for harmonics analysis has outperformance characteristics. [ABSTRACT FROM AUTHOR]

Identifying structural hole spanners that benefit from acting as bridges between communities is a core study in social network analysis. Existing methods for identification mainly focus on measuring the ability of users to control information propagation by bridging holes, while ignoring the impact of reinforcement of the holes themselves on the benefits of bridging spanners. A recent sociological study shows that the more reinforced a hole is, the more likely it is to bring high benefits to its spanners. In this paper, we propose a node embedding-based method ReHSe for identifying reinforced structural hole spanners in social networks. Specifically, an integrated embedding method is devised to extract features encoding reinforcement properties of nodes into a low-dimensional space. Further, to improve the robustness and accuracy of identification, an incremental learning strategy based on a reserved set is employed to train a scoring network in this subspace, to find top- k reinforced hole spanners. Extensive experimental results show that the performance of hole spanners identified by the proposed method outperforms several existing methods. [ABSTRACT FROM AUTHOR]

Alzheimer's disease (A.D.) is the most widespread type of Dementia, and it is not a curable neurodegenerative disease that affects millions of older people. Researchers were able to use their understanding of Alzheimer's disease risk variables to develop enrichment processes for longitudinal imaging studies. Using this method, they reduced their sample size and study time. This paper describes the primitive detective of Alzheimer's diseases using Neuroimaging techniques. Several preprocessing methods were used to ensure that the dataset was ready for subsequent feature extraction and categorization. The noise was reduced by converting and averaging many scan frames from real to DCT space. Both sides of the averaged image were filtered and combined into a single shot after being converted to real space. InceptionV3 and DenseNet201 are two pre-trained models used in the suggested model. The PCA approach was used to select the traits, and the resulting explained variance ratio was 0.99The Simons Foundation Autism Research Initiative (SFARI)—Simon's Simplex Collection (SSC)—and UCI machine learning datasets showed that our method is faster and more successful at identifying complete long-risk patterns when compared to existing methods. [ABSTRACT FROM AUTHOR]

Aiming at the inconsistency of manual detection of mobile phone screen defects, the image feature extraction of traditional machine learning is often set based on experience, resulting in unsatisfactory detection results. Therefore, a mobile phone screen defect detection model (Ghostbackbone) which is proposed by this paper based on YOLOv5 s and Ghostbottleneck. The bottleneck of Ghostbackbone mainly uses and improves the Ghostbottleneck of GhostNet. The attention module of Ghostbackbone uses Coordinated Attention and Depthwise Separable Convolution for parameter reduction. Finally, Ghostbackbone uses YOLOv5 as the object detector to train the mobile phone screen defect dataset. The experimental results show that the parameter quantity of Ghostbackbone is 24% of that of YOLOv5 s, the average time of detecting a single picture is only 2% lower than that of YOLOv5 s, and the mAP0.5 : 0.95 is 2% higher than that of MobilenetV3 s. [ABSTRACT FROM AUTHOR]

A drive-by download is a method of hackers planting the Web Trojan, which exploits browser vulnerabilities to execute malicious software. Because people usually access web pages with various browsers daily, drive-by downloads have become one of the most common threats in recent years. Most previous studies utilize the abstract syntax tree(AST) with deep learning methods to detect such attacks, which achieved high accuracy but are time-consuming and challenging to explain. Also, some methods use dynamic analysis, which needs a specific environment and is time-consuming with the complex operation. In order to solve these problems, the paper proposes DDIML, an explainable machine learning model based on novel features with static analysis. These features are extracted from five aspects: code obfuscation, URL redirection, special behaviors, encoding characters, and CSS attributes. The most popular machine learning algorithm, Random forest, is applied for building the classifier detection model. In addition, we use both local and global explanations to improve the model and prove that the proposed model could be trusted. The Experimental results show that our proposed model can efficiently detect drive-by downloads with a detection precision of 0.983 and a recall of 0.980. The average detection time for each sample is only 16.07ms in total. [ABSTRACT FROM AUTHOR]

Sports athletes not only exercise fast, but also suffer from the surrounding complex environment. Therefore, the video needs to be sequenced to improve processing efficiency. From the perspective of machine learning, this paper designs a spatial feature extractor based on CNN to extract time series features of sports. Moreover, this paper uses the support vector machine as the basis of the construction model to construct a feature extraction model based on support vector machine and random forest based on different situations. At the same time, this paper collects test data through the sports database and uses the swimming project as an example to analyze the model performance. Finally, the paper verifies the validity of the model by comparing and verifying methods. The research indicates that the proposed method has certain effectiveness and can provide theoretical reference for subsequent related research. [ABSTRACT FROM AUTHOR]

Hyperspectral Image (HSI) is usually composed of hundreds of capturing wavelength bands, which not only increase the size of the HSI rapidly but also impose various obstacles in classifying the objects accurately. Moreover, the traditional machine learning schemes utilize only the spectral features for HSI classification, which, therefore, neglect the spatial features that have a significant impact on the classification improvement. To address the aforementioned issues, in this paper, we propose to employ the principal component analysis (PCA), the baseline feature extraction method, and a thoughtfully designed stacked autoencoder, a deep learning-based feature extraction approach, for reducing the high dimensionality of the HSI and then propose a novel lightweight 3D-2D convolutional neural network (CNN) framework to concurrently exploit both spatial and spectral features from the dimensionality-reduced HSI for classification. In particular, PCA and stacked autoencoder are applied to reduce the high dimensionality of the original HSI and then the proposed 3D-2D CNN provides a combination of 3D and 2D convolution operations to extract the subtle spatial and spectral features for efficient classification. We well-adjust the proposed 3D-2D CNN architecture, and perform extensive experiments on three benchmark HSI datasets and compare our approach with the state-of-the-art classical and deep learning methods. Experimental results illustrate that we have achieved an overall accuracy of 99.73%, 99.90%, and 99.32% on Indian Pines, Pavia University, and Kennedy Space Center datasets, respectively, which outperform the classical machine learning and independent 2D and 3D CNN-based state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Tyre condition monitoring system (TCMS) is an emerging electronic safety system present in most of the new generation vehicles. It is intended to warn the driver when wheels are unbalanced or inflation pressure in a tyre goes below a certain recommended value. In India only limited vehicles having this critical safety system installed because it is highly expensive. The existing TCMS monitor only the tyre pressure using direct or indirect methods. In recent days the indirect TCMS using vibration signals and machine learning techniques are gaining momentum. This paper reports the effect of wheel balancing on a proposed affordable indirect TCMS. The present study goes one step forward to include the wheel balancing in the TCMS, thus increasing the scope of TCMS. The study was carried out using vibration signal data acquired from a low cost accelerometer placed on the wheel hub. The features required were taken out from the acquired data with the help of statistical feature extraction techniques. Features selection and feature classification were processed with J48 decision tree algorithm. The effect of wheel balancing in classification accuracy is clearly explained. [ABSTRACT FROM AUTHOR]

This paper presents a computational model for the unsupervised authorship attribution task based on a traditional machine learning scheme. An improvement over the state of the art is achieved by comparing different feature selection methods on the PAN17 author clustering dataset. To achieve this improvement, specific pre-processing and features extraction methods were proposed, such as a method to separate tokens by type to assign them to only one category. Similarly, special characters are used as part of the punctuation marks to improve the result obtained when applying typed character n-grams. The Weighted cosine similarity measure is applied to improve the B3 F-score by reducing the vector values where attributes are exclusive. This measure is used to define distances between documents, which later are occupied by the clustering algorithm to perform authorship attribution. [ABSTRACT FROM AUTHOR]

Fuzzy English recognition is affected by many factors, which leads to certain accuracy problems in intelligent recognition results. In order to improve the automatic recognition efficiency of fuzzy English, based on machine learning technology, this study constructs a neural network model. At the same time, this paper analyzes the research status and existing problems of handwritten character recognition, analyzes the model, and adopts multiple modules for automatic English recognition. In addition, the system is built on the basis of algorithms and model support, which makes fuzzy English recognition intelligent. Finally, in order to study the algorithm and model performance, the fuzzy English recognition is carried out through experiments. The research shows that the model constructed in this paper has certain recognition effect, which can be applied to practice, and can provide theoretical reference for subsequent related research. [ABSTRACT FROM AUTHOR]

Since small objects occupy less pixels in the image and are difficult to recognize. Small object detection has always been a research difficulty in the field of computer vision. Aiming at the problems of low sensitivity and poor detection performance of YOLOv3 for small objects. AFYOLO, which is more sensitive to small objects detection was proposed in this paper. Firstly, the DenseNet module is introduced into the low-level layers of backbone to enhance the transmission ability of objects information. At the same time, a new mechanism combining channel attention and spatial attention is introduced to improve the feature extraction ability of the backbone. Secondly, a new feature pyramid network (FPN) is proposed to better obtain the features of small objects. Finally, ablation studies on ImageNet classification task and MS-COCO object detection task verify the effectiveness of the proposed attention module and FPN. The results on Wider Face datasets show that the AP of the proposed method is 11.89%higher than that of YOLOv3 and 8.59%higher than that of YOLOv4. All of results show that AFYOLO has better ability for small object detection. [ABSTRACT FROM AUTHOR]

BACKGROUND: Breast cancer is a common disease in women. Early detection and early treatment can reduce breast cancer mortality. Studies have shown that breast cancer microcalcifications is one of the important clinical manifestations of early breast cancer, and sometimes even the only manifestation. When the mammography image shows typical malignant microcalcification, it can be diagnosed as breast cancer without any other signs of malignancy. In the aided diagnosis of microcalcifications, it is a crucial step to automatically find and locate regions of interest containing microcalcifications. However, the existing feature extraction method for microcalcifications only extracts features in the time domain or wavelet domain, and does not completely represent all the information of the region of interest. An extraction method based on the combination of Dual-Tree Complex Wavelet Transform (DTCWT) and texture features is proposed in the paper. METHODS: First, the processing operations including denoising, enhancement, and edge detection were performed on mammograms. Sub-image segmentation is then performed. DTCWT features and texture features are extracted for each sub-image.DTCWT features are combined with texture features, and then genetic algorithm is used for feature optimization. The features are classified by the Extreme Learning Machine (ELM) to achieve rapid detection and automatic extraction of ROI with microcalcifications. The experimental results verify that the feature model proposed in this paper has the highest detection rate for ROI regions. The ROI region extracted by the proposed feature model was used as subsequent experimental data. Three different methods were used to detect the microcalcifications, including Top-hat, wavelet transform, and methods combining Top-Hat and wavelet transform. RESULTS: The method was applied to 100 mammograms from the mammograms database of women in Northeast China. In the automatic extraction of ROI, the accuracy, sensitivity, specificity, positive accuracy and negative accuracy of the proposed model combined with DTCWT were 95.92%, 96.71%, 92.20%, 93.65%, 96.33%, respectively. When the Top-hat algorithm was used for microcalcifications detection, the sensitivity reached 89.6%, and the false positive detection rate was 2.6. When the wavelet transform algorithm was used for microcalcifications detection, the sensitivity was 91.1%, and the false positive detection rate was 3.28. When the combined algorithm was used for microcalcifications detection, the sensitivity was 86.7%, and the false positive detection rate decreased to 1.35. CONCLUSIONS: The proposed model combined with DTCWT features achieves better result in the automatic extraction of ROI. Moreover, in the subsequent detection of microcalcifications based on three methods, the three methods achieved better results in sensitivity and false positive detection rate, respectively. [ABSTRACT FROM AUTHOR]

At present, artificial intelligence for sports static image recognition is mostly in the action judgment stage, but less analysis on the action detail stage. Based on this, based on machine learning, this study uses static images and video sequences as carriers to improve traditional algorithm research and to perform motion gesture recognition. Through performance analysis, this paper explores the traditional algorithm and uses parameter analysis to improve the feature extraction and classification of traditional algorithms. Moreover, this paper uses the multi-scale feature approximation calculation method to improve the speed of the algorithm to extract features, and the algorithm is tested using the UCF motion data set and the self-created motion data set. In addition, this paper obtains representative motion video through data collection to test the effectiveness of the proposed algorithm. The research shows that the proposed algorithm has good performance and can provide theoretical reference for subsequent related research. [ABSTRACT FROM AUTHOR]

The World-Wide Web technology has become an indispensable part in human's life for almost all activities. On the other hand, the trend of cyberattacks is on the rise in today's modern Web-driven world. Therefore, effective countermeasures for the analysis and detection of malicious websites is crucial to combat the rising threats to the cyber world security. In this paper, we systematically reviewed the state-of-the-art techniques and identified a total of about 230 features of malicious websites, which are classified as internal and external features. Moreover, we developed a toolkit for the analysis and modeling of malicious websites. The toolkit has implemented several types of feature extraction methods and machine learning algorithms, which can be used to analyze and compare different approaches to detect malicious URLs. Moreover, the toolkit incorporates several other options such as feature selection and imbalanced learning with flexibility to be extended to include more functionality and generalization capabilities. Moreover, some use cases are demonstrated for different datasets. [ABSTRACT FROM AUTHOR]

With the development of the Internet, the rise of e-commerce has changed the shopping habits of most people. The research of this article is mainly divided into three parts. The first part is the theoretical foundation and core concept research. The second part of this article is a detailed method of establishing a predictive model based on machine learning image algorithms. In addition to reclassifying features, image algorithms are also used to optimize the model structure. The third part is the experimental results and analysis. After comparing with BP neural network and RBF neural network, through data analysis, the prediction model in this paper greatly improves the prediction accuracy and time, and the overall performance has a breakthrough. [ABSTRACT FROM AUTHOR]

This paper proposes a Wi-Fi-based indoor human detection system using a deep convolutional neural network. The system detects different human states in various situations, including different environments and propagation paths. The main improvements proposed by the system is that there is no cameras overhead and no sensors are mounted. This system captures useful amplitude information from the channel state information and converts this information into an image-like two-dimensional matrix. Next, the two-dimensional matrix is used as an input to a deep convolutional neural network (CNN) to distinguish human states. In this work, a deep residual network (ResNet) architecture is used to perform human state classification with hierarchical topological feature extraction. Several combinations of datasets for different environments and propagation paths are used in this study. ResNet's powerful inference simplifies feature extraction and improves the accuracy of human state classification. The experimental results show that the fine-tuned ResNet-18 model has good performance in indoor human detection, including people not present, people still, and people moving. Compared with traditional machine learning using handcrafted features, this method is simple and effective. [ABSTRACT FROM AUTHOR]

Named entity recognition is a fundamental task of natural language processing. The number of biomedical named entities is huge, the naming rules are not uniform, and the entity word formation is complex, which brings great difficulties to the biomedical named entity recognition. Traditional machine learning algorithms rely heavily on manual extraction of features. The quality of feature extraction directly affects the accuracy of entity recognition. In the biomedical domain, the cost of manually extracting features and annotating data sets is enormous. In recent years, deep learning methods that do not rely on artificial features have made great progress in many domains. This paper proposes a model based on Glove-BLSTM-CRF for biomedical named entity recognition. Firstly, the Glove model is used to train word vector with semantic features, and BLSTM is used to train word vector with character morphological features. The two are combined as the final representation of the word, then input into the BLSTM-CRF deep learning model to recognize the entity categories. The experimental results show that the model has achieved a better result in the JNLPBA 2004 biomedical named entity recognition task without relying on any artificial features and rules, and the F1 value reaches 75.62%. [ABSTRACT FROM AUTHOR]

Background: The quantitative features of a capnogram signal are important clinical metrics in assessing pulmonary function. However, these features should be quantified from the regular (artefact-free) segments of the capnogram waveform.Objective: This paper presents a machine learning-based approach for the automatic classification of regular and irregular capnogram segments.Methods: Herein, we proposed four time- and two frequency-domain features experimented with the support vector machine classifier through ten-fold cross-validation. MATLAB simulation was conducted on 100 regular and 100 irregular 15 s capnogram segments. Analysis of variance was performed to investigate the significance of the proposed features. Pearson's correlation was utilized to select the relatively most substantial ones, namely variance and the area under normalized magnitude spectrum. Classification performance, using these features, was evaluated against two feature sets in which either time- or frequency-domain features only were employed.Results: Results showed a classification accuracy of 86.5%, which outperformed the other cases by an average of 5.5%. The achieved specificity, sensitivity, and precision were 84%, 89% and 86.51%, respectively. The average execution time for feature extraction and classification per segment is only 36 ms.Conclusion: The proposed approach can be integrated with capnography devices for real-time capnogram-based respiratory assessment. However, further research is recommended to enhance the classification performance. [ABSTRACT FROM AUTHOR]

Link prediction tremendously gained interest in the field of machine learning and data mining due to its real world applicability on various fields such as in social network analysis, biomedicine, e-commerce, scientific community, etc. Several link prediction methods have been developed which mainly focuses on the topological features of the network structure, to figure out the link prediction problem. Here, the main aim of this paper is to perform feature extraction from the given real time complex network using subgraph extraction technique and labeling of the vertices in the subgraph according to the distance from the vertex associated with each target link. This proposed model helps to learn the topological pattern from the extracted subgraph without using the topological properties of each vertex. The Geodesic distance measure is used in labeling of the vertices in the subgraph. The feature extraction is carried out with different size of the subgraph as K = 10and K = 15. Then the features are fit to different machine learning classification model. For the evaluation purpose, area under the ROC curve (AUC) metric is used. Further, comparative analysis of the existing link prediction methods is performed to have a clear picture of their variability in the performance of each network. Later, the experimental results obtained from different machine learning classifiers based on AUC metric have been presented. From the analysis, we can conclude that AdaBoost, Adaptive Logistic Regression, Bagging and Random forest maintain great performance comparatively on all the network. Finally, comparative analysis has been carried out between some best existing methods, and four best classification models, to make visible that link prediction based on classification models works well across several varieties of complex networks and solve the link prediction problem with superior performance and with robustness. [ABSTRACT FROM AUTHOR]

Breast cancer is one of the major causes of women death worldwide. WHO organization has reported that 1 in every 12 women could be subjected to a breast abnormality during her lifetime. To increase survival rates, it is found that early detection of breast tumor is very critical. Mammography-based breast cancer screening is the leading technology to achieve this aim. However, it still can not deal with the cases where the tumor size less than 2mm. Thermography-based breast cancer detection methods can address this problem. In this paper, a breast cancer detection method is proposed. The proposed method is consists of four phases: (1) Image Pre-processing using homomorphic filtering, (2) Region of interest (ROI) Segmentation using K-mean clustering, (3) feature extraction using signature boundary, and (4) classification using Extreme Learning Machine (ELM). The proposed method is evaluated using the public dataset DMR-IR. Different activation functions in ELM are evaluated. The obtained results founded that "Tribas" is the best activation function under different experiments. It produced an accuracy result of 95.94% while talking 0.0469 second to detect the existence of malignant tumor, benign tumor or normal image. These promising results would be useful to develop thermography-based breast cancer detection system. [ABSTRACT FROM AUTHOR]

Reinforcement Learning (RL) is an effective approach to solve sequential decision making problems when the environment is equipped with a reward function to evaluate the agent's actions. However, there are several domains in which a reward function is not available and difficult to estimate. When samples of expert agents are available, Inverse Reinforcement Learning (IRL) allows recovering a reward function that explains the demonstrated behavior. Most of the classic IRL methods, in addition to expert's demonstrations, require sampling the environment to evaluate each reward function, that, in turn, is built starting from a set of engineered features. This paper is about a novel model-free IRL approach that does not require to specify a function space where to search for the expert's reward function. Leveraging on the fact that the policy gradient needs to be zero for an optimal policy, the algorithm generates an approximation space for the reward function, in which a reward is singled out employing a second-order criterion. After introducing our approach for finite domains, we extend it to continuous ones. The empirical results, on both finite and continuous domains, show that the reward function recovered by our algorithm allows learning policies that outperform those obtained with the true reward function, in terms of learning speed. [ABSTRACT FROM AUTHOR]

In order to pre-warning the product quality risk of the e-commerce platform, this paper studies the machine learning algorithm for the products quality risk assessment, which propose the Fuzzy C-Means clustering algorithm for the feature extraction and the Cost Sensitive Leaning (CSL)-Naive Bayesian algorithm to construct the assessment model for E-commerce product quality risk form the massive and unbalanced data. The experimental results show that the Machine Learning algorithm based on Spark has better scalability and superiority in the large-scale data environment, which can accurately identify e-commerce product quality risk. [ABSTRACT FROM AUTHOR]

Human activity recognition emerges as one of the prominent research areas in the recent past. However, the activity recognition still encounters many challenges like reliability of sensor data and accuracy of prediction that severely affects the aspect of decision making. In this paper, a futuristic framework has been proposed and experimented to build a precision-centric activity recognition method by analyzing the data obtained from Environment Monitoring System (EMS) and Personalized Positions Detection System (PPDS) using machine learning methods such as AdaBoost, Support Vector Machine (SVM) and Probabilistic Neural Networks (PNN). Further, the proposed approach utilizes the Dempster-Shafer Theory (DST)-based complete sensor data fusion thereby improving the global activity recognition performance. Finally, the proposed approach is validated using a real-world dataset obtained from UCI machine learning repository. The results conclude that the proposed activity recognition framework outperforms its existing context/situation-awareness approaches in terms of reliability, efficiency, and accuracy. [ABSTRACT FROM AUTHOR]

The ever increasing advances in the field of biotechnology and health information sciences have led to large electronic health records (EHRs), which in turn contains important genetic and clinical information. Machine learning and data mining techniques are playing vital and indispensible efforts to intelligently convert available data into useful information for effective medical diagnosis. But, designing effective prediction and diagnosis techniques for diabetes mellitus (DM) are getting more attention than ever before. Thus, a novel fuzzy rule miner (ANT_FDCSM) derived from ant colony meta-heuristic for diagnosis of diabetic patients has been proposed in this paper. A few important improvements have been suggested to improve the performance of traditional ant colony optimization induced decision tree classifier. The first improvement is done to optimize search space of construction graph by employing a novel approach for optimal split point selection. Secondly, to compute heuristic information, a hybrid node split measure (SW_FDCSM) is presented. SW_FDCSM is a combination of attribute significance weight (SW) with a new fuzzy variant (Fuzzy_DCSM) of famous distinct class split measure (DCSM). The improvements have been proposed to generate comprehensive rule set while maintaining good accuracy, sensitivity and specificity count. A 10 fold cross validation (10-FNo) is applied on Pima Indian Diabetes (PID) data set to validate the performance of the proposed fuzzy rule miner (ANT_FDCSM). [ABSTRACT FROM AUTHOR]

Object tracking is an efficient technique adopted in video surveillance applications for monitoring a particular object in a zone. This paper presents a novel efficient tracking and counting approach for Pedestrians in video sequences. For object detection, a Gaussian Mixture Model (GMM) is used to obtain binary masks. In the Speeded Up Robust Feature feature recognition, only the features of the object are retained. This significantly improves the precision of the Speeded Up Robust Feature method. For clustering the features, an enhanced grouping algorithm Density based Spatial Clustering of Application with Noise is proposed in which the motion features are grouped and the remaining features are excluded. The features are tracked based on optical flow method. For counting the number of objects, the Pedestrian eigen vectors are created based on the Speeded Up Robust Features and the eigen vectors are trained with a SVM (support vector regression machine). The proposed work combines the object detection, feature extraction, and objects counting. The experimental results validate that the proposed pedestrian tracking and counting method is efficient than the existing approaches. [ABSTRACT FROM AUTHOR]

With the increasingly growing internal and external attacks on computer systems and online services, cybersecurity has become a vibrant research area. Countering intrusive attacks is a daunting task with no universal magic solution that can successfully handle all scenarios. A variety of machine-learning and computational intelligence techniques have been applied extensively to detect and classify these attacks. However, the effectiveness of these techniques greatly depends on the adopted data preprocessing methods for feature extraction and engineering. This paper presents an extended taxonomy of the work related to intrusion detection and reviews the state-of-the-art techniques for data preprocessing. It offers a critical up-to-date survey which can be an instrumental pedagogy to help junior researchers conceive the vast amount of research work and gain a holistic view and awareness of various contemporary research directions in this domain. [ABSTRACT FROM AUTHOR]

EEG signal is considered a dynamical system, difficult and complex to learn. Therefore Brain Computer Interface Systems need to manage specific time variations of the EEG since the extracted feature are non-stationary. This paper presents a study to test Extreme Learning Machine as a suitable classification method for Motor Imagery Brain Computer Interface. In order to take in to account the time course of the signals new descriptors from three widely known Feature Extraction methods (Power Spectral Density, Hjorth parameters and Adaptive AutoregRessive coefficients) have been obtained by three different techniques: central window, averaging features and linking features. Results shows that these new descriptors have improved the performance of the Extreme Learning Machine with respect classical techniques. [ABSTRACT FROM AUTHOR]

Inspired by unsupervised feature learning and deep learning, this paper provides a novel classification method for advanced suspension system based on Deep Neural Networks (DNNs). Sparse autoencoder and softmax regression are chosen to form deep structure and the parameters are trained by deep learning. Aiming at showing the superiority of DNNs based road classification method, a simulation of a B-class vehicle with skyhook control is performed in CarSim, and three measurable system responses, i.e., centre of gravity (C.G.) of sprung mass acceleration, rattle space and unsprung mass acceleration are chosen and three independent classifiers are established. Simulation results show that the classifier using unsprung mass acceleration has the highest accuracy and better performance than existing methods. Because of the adaptive learning ability and the deep structure, the proposed method can save work and provide higher classification accuracy. [ABSTRACT FROM AUTHOR]

Magnetic Resonance Imaging (MRI) is a diagnostic tool of remarkable potential in the area of neuroscience and clinical neuroimaging. The diagnostic accuracy can be limited by incompetence of the operating personnel, which can be supplemented by machine learning algorithms for classification of physiology and pathology. This paper uses effective information feature extraction, principal component analysis (PCA) for feature reduction and support vector machine (SVM) for classification of multi-sequence MR images of 7 patients. All axial slices of the brain are classified into normal and abnormal images. Various methods for feature extraction were tested among which effective information yielded the highest accuracy of 80.8% in a set of 677 images used for training and testing. The sensitivity and specificity were 80% and 81.06%, respectively. Different grid sizes were tested, and the highest accuracy was reported for 2 × 2 which indicates that the feature extraction must be taken over a small grid to ensure detection of minor variation from normal. The image sequences tested considered in the study are T1 weighted, T2 weighted, Fluid-attenuated inversion recovery (FLAIR), and post contrast T1 weighted. T2 weighted images were best classified with the maximum accuracy of 95.97%. This method proved to be effective to classify the images of all four sequences with accuracy ranging from 92-96%. The method was also tested with out of sample data and the accuracy obtained was 72.4%. The novelty of this work lies in the classification of multi-sequential images using all the different slices of the patient which includes the top of the skull as well as the mandible. The slices differ significantly as the spread of the tumor varies with each slice. The slices are taken at 5mm gap and the tumor can have a thickness less or more than the slice gap considered for the scan. [ABSTRACT FROM AUTHOR]

In recent years the number of new malware threats has increased significantly, causing a damage of billions of dollars globally. To counter this aggressive malware attack, the anti-malware industry needs to be able to correctly classify malware in order to provide defense against them. Consequently, malware classification has been an active area of research, and a multitude of malware classification approaches have been proposed in the literature. This paper evaluates two methods of sequence classification based on Hidden Markov Model, namely the maximum likelihood and similarity-based methods, for classification of malware using a large and comprehensive dataset. System calls generated by known malware during execution are used as observation sequences to train the Hidden Markov Models. Malware samples are evaluated against the trained models to produce similarity vectors, which are used in the maximum likelihood and similarity-based classification schemes to predict the family for an unknown malware sample. Comparison of the two schemes shows that combining the powerful statistical pattern analysis capability of Hidden Markov Models and discriminative classifiers in the similaritybased method results in a significantly better classification performance as compared to the maximum likelihood approach. Furthermore, evaluation of different classifiers in the similarity-based method demonstrates that Random Forest classifier performs better than other classifiers on malware similarity vectors. [ABSTRACT FROM AUTHOR]

Steganalysis analyzes the existence of embedded secret information in cover. It can be divided into two categories, special steganalysis and blind steganalysis, according to the feature extraction and classification. Steganalysis can be assigned to classification problems, which based on the training set to determine the suspicious cover's class label. Essentially, it is two-class problem in pattern recognition or machine learning. Selecting the right feature for classification is very important and difficult. There has been little research that deals with the feature selection and feature extraction problem with specific respect to steganalysis. This paper studies the influence of selected feature to the steganalysis. It is crucial that selected features are very sensitive to the embedding changes, but insensitive to the image content. First, the basic framework is described for image steganalysis, which includes five parts: training/testing images set, feature search/selection, feature extraction/feature vectors, training classifier, training model and classifier parameter estimator. We then classify the existing feature according to the domain which belonging to. Finally, we do the experiment to compare the performance of different feature by use different classifier, such as ANN and SVM. Through our experiment, although we use small data set, but we find out the optimized features for classification. [ABSTRACT FROM AUTHOR]

This paper proposes a novel approach for medical x-ray image classification using fuzzification of Contourlet-based Center Symmetric Local Binary Patterns (CCS-LBPs). The proposed classification method consists of three stages. In the first stage, local features are obtained by partitioning each image into 25 overlapping sub-images, computing the 2-level contourlet transform of each subimage and extracting CS-LBPs from each resulting subband. In the second stage, fuzzy logic using reduced CCS-LBPs is employed to determine the degree of membership of subimages to each class. Finally, in order to assign images to their respective classes, we utilize membership values as the input of classifiers such as support vector machine (SVM) and k-nearest neighbor (K-NN). This work makes a major contribution to improve the performance of these classifiers. We conducted experiments on a subset of IRMA dataset to evaluate the effectiveness of our classification scheme. Experimental results reveal that the proposed scheme not only achieves a very good performance but also learns well even with a small number of training images. [ABSTRACT FROM AUTHOR]

In this paper, we introduce an incremental version of recently proposed constrained Linear Discriminant Analysis (LDA). In addition of application in constrained LDA problems, our algorithm which we call Online Discriminative Component Analysis (ODCA) is usable in standard incremental LDA problems. ODCA incrementally computes the solution of LDA with the time complexity lower than most incremental algorithm for LDA while keeps the accuracy of final result as close as possible to offline version. This is done using a special formulation for the scatter matrix updating along with Eigen-space calculation. By exploiting such formulation, the proposed algorithm made capable of updating the solution where a data point added or removed from the problem. It is also usable in problems where its data points have concept drift property. To show efficiency of proposed algorithm, its speed is compared to other existing incremental algorithms as order of complexity. In addition, the classification accuracy of our approach is experimentally compared to other algorithms. [ABSTRACT FROM AUTHOR]