Improving Data Classification Research Articles

Evidential dissonance measure in robust multi-view classification to resist adversarial attack

Multi-view learning is effective in improving data classification accuracy through integrating information from multiple sources. To guarantee the reliability of multi-view classification, the trusted multi-view classification methods have been explored. However, extant trusted multi-view classification methods are still vulnerable to the low-quality views containing adversarial samples. This vulnerability arises from a challenge in accurately assessing the quality of data views that are subjected to adversarial attacks. To address this issue, we propose a robust multi-view classification method with dissonance measure of adversarial samples. Specifically, the proposed method utilizes the evidential dissonance measure in subjective logic to assess the quality of data views when encountering adversarial attacks. Based on the dissonance measure, we further propose a strategy of dissonance-aware belief integration for multi-view information fusion and construct the inter-view evidential gradient penalty in the learning objective of multi-view classification to improve the model robustness to resist adversarial samples. Experiments on diverse multi-view datasets confirm the reliability and robustness of the proposed multi-view classification method.

Improve data classification performance in diagnosing diabetes using the Binary Exchange Market Algorithm

Today's lifestyle has led to a significant increase in referrals to medical centers to diagnose various diseases. To this end, over the past few years, researchers have turned to new diagnostic methods, including data mining and artificial intelligence, intending to facilitate the detection process and increase reliability. The high volume of data available in medical centers can be considered one of the main problems in using these methods. The optimal selection of essential and influential features reduces the maximum dimension for better diagnosis with more reliability of results. In this paper, a new approach uses a Binary Exchange Market Algorithm (BEMA) to identify essential and practical features in the diabetes dataset and determine the best algorithm binary function (type of sigmoid function) to improve the performance of the EMA algorithm is presented. For validation and efficiency of the proposed BEMA algorithm, several SVM, KNN, and NB classification models have been used to train and test the final model. The results obtained from the evaluations show that the proposed BEMA-SVM combined method has a better performance than the previous methods to improve accuracy and offer an effect equivalent to 98.502%. Also, to provide better results and more reliability than the proposed method, researchers can use a combination of several classes with the proposed method, which is outside the scope of this study.

A hybrid approach for improving data classification based on PCA and enhanced ELM

A hybrid approach for improving data classification based on PCA and enhanced ELM

Hybridization of Neutrosophic Logic with Quasi-Oppositional Chimp Optimization based Data Classification Model

Data classification is the procedure of investigating structured or unstructured data and forming it into distinct classes depending upon file types, size, etc. It assist the organizations to derive important solutions based on the data and helps decision making process. The computational intelligence techniques such as neural computing, fuzzy logic, machine learning, etc. can be used to design effective data classification models. This study offers a Hybridization of Neutrosophic Logic with Quasi-Oppositional Chimp Optimization based Data Classification (HNLQOCO) model. The presented HNLQOCO algorithm aims to integrate the concepts of NL and QOCO algorithm for improved data classification outcomes. Besides, the QOCO algorithm is designed by incorporating the concepts of quasi oppositional based learning (QOBL) with traditional chimp optimization algorithm (COA). Here, the NL is applied to represent various kinds of knowledge and the QOCO algorithm is applied to tune the produced NS rules. The experimental result analysis of the HNLQOCO model is tested using three benchmark medical dataset. The obtained results reported the significant performance of the HNLQOCO model over the other methods.

The development of real-time digital PCR technology using an improved data classification method

For digital polymerase chain reaction (PCR), data classification is always a crucial task. The dynamic real-time amplification process information of each partition is always ignored in typical digital PCR analysis, which can easily lead to inaccurate outcomes. In this work, an integrated device that offers real-time chip-based digital PCR analysis was established. In addition, an enhanced process-based classification model (PAM) was built and trained. And then the device and the analytical model were employed in classification tasks for different concentrations of Epstein-Barr Virus (EBV) plasmid quantification assays. The results indicated that the real-time analysis device achieved a linearity of 0.97, the classification method was able to distinguish the false-positive curves, and the recognition error of positive wells was decreased by 64.4% compared with typical static analysis techniques when low concentrations of samples were tested. With these advantages, it is supposed that the real-time digital PCR analysis apparatus and the improved classification method can be employed to enhance the performance of digital PCR technology.

Novel Robust Augmentation Approach Based on Sensing Features for Data Classification

Learning networks have become extremely powerful tools for data classification. One drawback in many of these applications, however, is that they require a large dataset for efficient performance. Medical imaging, for example, cannot make large enough databases available due to strong data privacy concerns and high data collection costs. Fortunately, data augmentation can provide a solution to improve the classification performance of these systems. This paper proposes a novel method for data augmentation based on the compressive sensing technique. Our approach uses sensing matrices to manipulate data features into different spaces in a compressed form, which are then combined and used as extensions of the original limited dataset. This new feature representation is allowing for a more robust learning network that thereby enables improved data classification. When applying our approach to binary-class and multi-class datasets, results showed that our approach can significantly improve the classifier performance and enhance classification results for both datasets.

A Machine Learning Approach to Road Surface Anomaly Assessment Using Smartphone Sensors

Road surface quality is essential for improving driving experience and reducing traffic accidents. Traditional road condition monitoring systems are limited in their temporal (speed) and spatial (coverage) responses needed for maintaining overall road quality. Several alternative systems have been proposed that utilize sensors mounted on vehicles. In particular, with the ubiquitous use of smartphones for navigation, smartphone-based road condition assessment has emerged as a promising new approach. In this paper, we propose to analyze different multiclass supervised machine learning techniques to effectively classify road surface conditions using accelerometer, gyroscope and GPS data collected from smartphones. Our work focuses on classification of three main class labels- smooth road, potholes, and deep transverse cracks. We hypothesize that using features from all three axes of the sensors provides more accurate results as compared to using features from only one axis. We also investigate the performance of deep neural networks to classify road conditions with and without explicit manual feature extraction. Our results indicate that models trained with features from all axes of the smartphone sensors outperform models that use only one axis. We also observe that the use of neural networks provides a significantly improved data classification. The machine learning approach discussed here can be implemented on a larger scale to monitor roads for defects that present a safety risk to commuters as well as to provide maintenance information to relevant authorities.

A Comprehensive Framework for Ontology Based Classifier using Unstructured Data

The knowledge contained within the natural language data can be used to build expert systems. Classifying unstructured data using ontology and text classification algorithms to extract information is one way of approaching the problem of building intelligent systems. One major problem with text processing is most data generated is unstructured and ambiguous, as, data with a structure helps to identify meaningful patterns and eventually exhibit the latent knowledge. Ambiguity in natural language affects accuracy of categorization. Also, Natural Language Processing techniques when combined with semantic data modeling through ontological knowledge will also solve the problem of domain knowledge representation thereby enabling improved data classification facilities, particularly in large datasets where number of features scale to unmanageable proportions. In this paper, the domain knowledge is presented as a knowledge graph, derived from the semantic data modeling. Further, to achieve better Multi Class classification, Multinomial Naive Bayes algorithm is applied to categorize items in their respective classes. For the experiments, Data about various news groups were used for testing the accuracy of the model. Experimental results have proved that the proposed classifier performs better compared to existing systems.

Improving data classification accuracy in sensor networks using hybrid outlier detection in HAR

Improving data classification accuracy in sensor networks using hybrid outlier detection in HAR

Legitimacy of Personal Data Mining and Related Regulatory Approaches: China’s Choices

Personal data mining is a crucial but controversial topic in the big data era. It is widely accepted in the technical filed as a big data technology. While more and more people begin to question its legitimacy for it causes many social and legal problems. This paper tries to demonstrate the legitimacy of personal data mining from the following reasons: a) personal data has certain public goods attribute and sharing characteristic, which is the internal premise of personal data mining; b) data appreciation theory is the theoretical basis of personal data mining; c) the principle of interests balance is the logical requirement of personal data mining; d) big data sharing economy provides practical soil for personal data mining. Based on the legitimacy analysis and China’s data development practices, personal data mining should be encouraged. Meanwhile, it is particularly important to regulate the approaches of personal data mining in four aspects: a) establishing the purpose of combining personal data mining and protection; b) balancing the interests of all parties in the process of personal data mining; c) improving data classification legislation; d) innovating technical means to achieve privacy protection.

Comparisons of different methods for balanced data classification under the discrete non-local total variational framework

Because balanced constraints can overcome the problems of trivial solutions of data classification via minimum cut method, many techniques with different balanced strategies have been proposed to improve data classification accuracy. However, their performances have not been compared comprehensively so far. In this paper, we investigate seven balanced classification methods under the discrete non-local total variational framework and compare their accuracy performances on graph. The two-class classification problem with equality constraints, inequality constraints and Ratio Cut, Normalized Cut, Cheeger Cut models are investigated. For cases of equality constraint, we firstly compare the Penalty Function Method (PFM) and the Augmented Lagrangian Method (ALM), which can transform the constrained problems into unconstrained ones, to show the advantages of ALM. The other cases are all solved using the ALM also. In order to make the comparison fairly, we solve all models using ALM method and using the same proportion of fidelity points and the same neighborhood size on graph. Experimental results demonstrate ALM with the equality balanced constraint has the best classification accuracy compared with other six constraints. 200 words.

A hybrid approach for improving data classification based on PCA and enhanced ELM

A hybrid approach for improving data classification based on PCA and enhanced ELM

P 136 Brain network analysis of EEG data in the service of clinical assessment – utilizing big data and prior theoretical knowledge to identify a biomarker for mTBI in adolscents

Mild traumatic brain injury (mTBI) in adolecents has gained increased attention in recent years amongst parents, clinicians, and researchers due to their growing rate and hazardous outcomes. Electroencephalography (EEG) and Event-Related Potential (ERP) have been encouraged as a diagnostic tool for mTBI due to its objectivity and cost-effectiveness. However, extracting clinically meaningful neuro-cognitive information from human EEG/ERP is challenging, particularly in highly variable groups like adolescents. Therefore, it is not surprising that despite them being especially susceptible to the effects of mTBI, no sensitive and specific application of EEG has yet been determined for this age group. To overcome these challenges, we have developed and applied a framework, termed Brain Network Activity (BNA) analysis, which utilizes a hybrid approach of very large datasets, novel signal processing technology, and academic knowledge (see Stern, Y., Reches, A., and Geva A. “Using Spatiotemporal Features in the Brain Network Activation Analysis for Improved Data Classification”. Frontiers in computational neuroscience, 2016). For the BNA core database, spatiotemporal data from 15,100 EEG recording files of healthy individuals performing various computerized cognitive tasks were extracted. Advanced machine learning and prior knowledge from the ERP literature was used to assemble an optimal set of brain network models for various cognitive functions, such as attention, memory, motor control, and sensory processing. This resulted in reference brain network models, to which the BNA of an individual or of a whole patient group can be compared to in terms of quantitative scores and qualitative brain map illustrations. In cases of specific sub-groups such as adolescents with mTBI, these BNA scores and maps may be applied to identify biomarkers specific to that population. In order to generate an mTBI biomarker for adolescents, BNA analysis was applied to EEG recordings of 107 healthy participants and 36 mTBI adolescent patients while they performed an auditory oddball task. The BNA features of all participants were extracted and included in repeated-measures analyses of variance (ANOVA) with EEG session gain (2nd vs 1st, 3rd vs 2nd), interval between sessions ( 2 months) and group (healthy, mTBI) as factors. Results showed a large negative evoked potential component in frontal-central regions between ∼200 and 500 ms following the standard stimulus for both groups (mTBI and healthy) during the baseline EEG recording sessions. The average negative amplitude decreased in the healthy control group’s 2nd EEG recording session. It remained steady in the concussed group’s 2nd (post-mTBI) session, and was significantly different from healthy controls. The pre- to post-mTBI gain in amplitude was correlated with the time interval between recording sessions (r = −0.47, p-value

A hierarchical fuzzy logic systems frame work for data security

Data security and privacy plays a key role in carrying out day to day and long term business operations successfully. Organizations need to protect their data in the areas of data security and privacy. Therefore data security and privacy is a significant activity for large organizations. One novel method used to achieve data security is to use fuzzy logic based data classification to determine the required data security levels based on various government and business policies. Classifying data provides organizations the facilities to categorise organisational data based on organisational policies. Organisational data needs to be evaluated. Fuzzy logic based categorisation is explored in this paper. The method employed is based on Mohammadian et al. (Int J Electron Financ 3(4):374–386, 2009). This paper considers the application of a hierarchical fuzzy logic (HFL) classification method for categorizing organisational data based on governmental and business data policies. This paper demonstrates the use of a HFL to improve data classification. HFL reduces the size of the required fuzzy knowledge base system of the HFL data classification. This paper demonstrates how a HFL system can be employed to improve data security, privacy and user access control to determine data and user access control based on data sensitivity government and organizational data security requirements and user level security measures.

Lexicon-corpus Based Korean Unknown Foreign Word Extraction and Updating Using Syllable Identification

This paper presents an efficient text mining method focusing on extraction and updating of unknown words (unknown foreign words) to improve data classification and POS tags. Proposed methods can also help to improve the accuracy of mining frequent pattern and association rules from unstructured (textual) data. Many researches have been done by numerous scholars on estimation and segmentation for unknown words, but, they are limited to grammatical and linguistic rules with limited vocabulary. In our project we have consider the fact, that no language is free from the influence of foreign languages, especially, country like Korea where there is a rapid improvement in the area of culture and media and the frequent usage of these foreign languages, resulted in mixing up different languages, their style along with slangs and also abbreviated words in daily life and conversation. The main characteristic of our system is to find such unknown foreign words and update them to appropriate words, which depends on available information through dictionaries. We have also explained the essential natural language processing (NLP) tools used for data processing. Our proposed method used simple but efficient techniques, first it converts the data into structured form, using data preprocessing techniques. In this phase data passes through different stages, such as, cleaning, integration and selection of important data, and then it gets organized into databases structure for further analysis and processing. This database consists of different kinds of dictionaries, our system heavily based on dictionaries. We have manually created various kinds of dictionaries for different kinds of unknown foreign words processing and analysis with the help of our team members. Our proposed methods for discovering and updating foreign unknown word, first discovers the foreign word using morphological analysis with the help of automatically and manually created dictionaries, then suffix trimming and word segmentation, next our algorithm checks for its different written pattern using dictionaries according to its spelling and synonym word in native language (Korean) and also, updates the POS tags. We have tested on different collection of data from economics news, beauty & fashion and college student blogs, the results have shown great efficiency and improvement, and they were adequate enough to research further.

Hybrid Pattern Recognition Using a Combination of Different Features

We propose a hybrid pattern recognition method that effectively combines two different features for improving data classification. We first extract the PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) features, both of which are widely used in pattern recognition, to construct a set of basic features, and then evaluate the separability of each basic feature. According to the results of evaluation, we select only the basic features that contain a large amount of discriminative information for construction of the combined features. The experimental results for the various data sets in the UCI machine learning repository show that using the proposed combined features give better recognition rates than when solely using the PCA or LDA features.

Calibrating animal-borne proximity loggers.

Summary Growing interest in the structure and dynamics of animal social networks has stimulated efforts to develop automated tracking technologies that can reliably record encounters in free‐ranging subjects. A particularly promising approach is the use of animal‐attached ‘proximity loggers’, which collect data on the incidence, duration and proximity of spatial associations through inter‐logger radio communication. While proximity logging is based on a straightforward physical principle – the attenuation of propagating radio waves with distance – calibrating systems for field deployment is challenging, since most study species roam across complex, heterogeneous environments.In this study, we calibrated a recently developed digital proximity‐logging system (‘Encounternet’) for deployment on a wild population of New Caledonian crows Corvus moneduloides. Our principal objective was to establish a quantitative model that enables robust post hoc estimation of logger‐to‐logger (and, hence, crow‐to‐crow) distances from logger‐recorded signal‐strength values. To achieve an accurate description of the radio communication between crow‐borne loggers, we conducted a calibration exercise that combines theoretical analyses, field experiments, statistical modelling, behavioural observations, and computer simulations.We show that, using signal‐strength information only, it is possible to assign crow encounters reliably to predefined distance classes, enabling powerful analyses of social dynamics. For example, raw data sets from field‐deployed loggers can be filtered at the analysis stage to include predominantly encounters where crows would have come to within a few metres of each other, and could therefore have socially learned new behaviours through direct observation. One of the main challenges for improving data classification further is the fact that crows – like most other study species – associate across a wide variety of habitats and behavioural contexts, with different signal‐attenuation properties.Our study demonstrates that well‐calibrated proximity‐logging systems can be used to chart social associations of free‐ranging animals over a range of biologically meaningful distances. At the same time, however, it highlights that considerable efforts are required to conduct study‐specific system calibrations that adequately account for the biological and technological complexities of field deployments. Although we report results from a particular case study, the basic rationale of our multi‐step calibration exercise applies to many other tracking systems and study species.

NÂNG CAO HIỆU QUẢ PHÂN LỚP DỮ LIỆU BẰNG CÁCH KẾT HỢP SINH THÊM PHẦN TỬ VỚI GIẢM SỐ CHIỀU THUỘC TÍNH DỰA TRÊN BÌNH PHƯƠNG TỐI THIỂU TỪNG PHẦN

NÂNG CAO HIỆU QUẢ PHÂN LỚP DỮ LIỆU BẰNG CÁCH KẾT HỢP SINH THÊM PHẦN TỬ VỚI GIẢM SỐ CHIỀU THUỘC TÍNH DỰA TRÊN BÌNH PHƯƠNG TỐI THIỂU TỪNG PHẦN

A Fuzzy Decision Tree for Processing Satellite Images and Landsat Data

Satellite and airborne images, including Landsat, ASTER, and Hyperspectral data, are widely used in remote sensing and Geo- graphic Information Systems (GIS) to understand natural earth related processes, climate change, and anthropogenic activity. The nature of this type of data is usually multi or hyperspectral with individual spectral bands stored in raster file structures of large size and global coverage. The elevated number of bands (on the order of 200 to 250 bands) requires data processing algorithms capable of extracting information content, removing redundancy. Conventional statistical methods have been devised to reduce dimension- ality however they lack specific processing to handle data diversity. Hence, in this paper we propose a new data analytic technique to classify these complex multidimensional data cubes. Here, we use a well-known database consisting of multi-spectral values of pixels from satellite images, where the classification is associated with the central pixel in each neighborhood. The goal of our proposed approach is to predict this classification based on the given multi-spectral values. To solve this classification problem, we propose an improved decision tree (DT) algorithm based on a fuzzy approach. More particularly, we introduce a new hybrid classification algorithm that utilizes the conventional decision tree algorithm enhanced with the fuzzy approach. We propose an improved data classification algorithm that utilizes the best of a decision tree and multi-criteria classification. To investigate and evaluate the performance of our proposed method against other DT classifiers, a comparative and analytical study is conducted on well-known Landsat data.

Discovery of medical Big Data analytics: Improving the prediction of traumatic brain injury survival rates by data mining Patient Informatics Processing Software Hybrid Hadoop Hive

Abstract Entering medical encounter data by hand is time-consuming. In addition, data are often not entered into the database in a timely enough fashion to enable their use for subsequent mission planning. The Patient Informatics Processing Software semi-automates the data collection process onboard ships. Then data within these images are captured and used to populate a database, after which multiple ship databases are used for reporting and analysis. In this paper, we used the Patient Informatics Processing Software Hybrid Hadoop Hive to orchestrate database processing via various ships, by marshaling the distributed servers, running the various tasks in parallel, managing all of the communications and data transfers between the various parts of the system, and providing for redundancy and fault tolerance. Then we employed the Apache Hive as a data warehouse infrastructure built on top of Hadoop for data summarization, query, and analysis to identify traumatic brain injury (TBI) as well as other injury cases. Finally, a proposed Misdiagnosis Minimization Approach method was used for data analysis. We collected data on three ship variables (Byrd, Boxer, Kearsage) and injuries to four body regions (head, torso, extremities, and abrasions) to determine how the set of collected variables relates to the body injuries. Two dimensions or canonical variables (survival vs. mortality) were necessary to understand the association between the two sets of variables. Our method improved data classification and showed that survival, mortality, and morbidity rates can be derived from the superset of Medical Operations data and used for future decision-making and planning. We suggest that an awareness of procedural errors as well as methods to reduce misclassification should be incorporated into all TBI clinical trials.