Data Mining - Concepts and Techniques.

What is Twitter

Anomaly detection is an important problem that has been well-studied within diverse research areas and application domains. The aim of this survey is two-fold, firstly we present a structured and comprehensive overview of research methods in deep learning-based anomaly detection. Furthermore, we review the adoption of these methods for anomaly across various application domains and assess their effectiveness. We have grouped state-of-the-art research techniques into different categories based on the underlying assumptions and approach adopted. Within each category we outline the basic anomaly detection technique, along with its variants and present key assumptions, to differentiate between normal and anomalous behavior. For each category, we present we also present the advantages and limitations and discuss the computational complexity of the techniques in real application domains. Finally, we outline open issues in research and challenges faced while adopting these techniques.

Deep Learning for Anomaly Detection: A Survey.

Emotion recognition from speech signals is an important but challenging component of Human-Computer Interaction (HCI). In the literature of speech emotion recognition (SER), many techniques have been utilized to extract emotions from signals, including many well-established speech analysis and classification techniques. Deep Learning techniques have been recently proposed as an alternative to traditional techniques in SER. This paper presents an overview of Deep Learning techniques and discusses some recent literature where these methods are utilized for speech-based emotion recognition. The review covers databases used, emotions extracted, contributions made toward speech emotion recognition and limitations related to it.

/pdf/speech-emotion-recognition-using-deep-learning-techniques-a-zl8wl0vs5t.pdf

Speech Emotion Recognition Using Deep Learning Techniques: A Review

Emotion recognition using deep learning approach from audio–visual emotional big data

People can post their comments on public events on the Internet, such as their ideas, emotions, and attitudes that can affect others. Online public opinions may affect the stability or security of the country because of the speed and convenience of information disseminating on the Internet. This article proposes the public opinion cellular automata for situation deduction to predict the possible trending of public events. In the cellular automata, online users are represented by cells, and their eigenvalues are calculated from the user’s historical comment data. The cell and their neighbors form a cellular space, whose topology is a directed graph. We set the state of each cell based on its attributes and initialize the cellular automata. The automata can deduce public opinion and predict the trend of public opinion by following the set evolutionary rules in advance. Experiments with the real data from Sina Weibo online users show that the cells in the cellular automata can accurately simulate users’ behavior, and the deduction results are close to the trend of real historical events. The cellular automata-based prediction of the number of participants and emotional trending are more accurate than other methods.

POCA4SD: A Public Opinion Cellular Automata for Situation Deduction

With the rise of video surveillance applications for analyzing real-time and batching video data in large scales, traditional video processing systems are being challenged due to real-time, intelligence and fault-tolerance demand for networked high resolution and large-scale video processing. These challenges can be further exacerbated by the existing predicaments of multi-platform, multi-format, multi-codec on video itself. The emergence of cloud computing and big data techniques makes it possible to manage complicated intelligent processing for large-scale video data. This paper proposes a general cloud-based video platform that can provide a robust solution to intelligent analytic and storage for video data, which is called ViCiBaF architecture (Video Cloud integrated with Batch processing and Fast processing) architecture. This ViCiBaF architecture is elaborated through an implementation that can effectively handle massive surveillance video data, where real-time analysis, batch processing, distributed storage and cloud services are seamlessly integrated to meet the requirements of intelligent analysis, real time, fault tolerance and massive storage for massive video data. The evaluations show that the proposed approach is efficient.

Towards a High Speed Video Cloud Based on Batch Processing Integrated with Fast Processing

To understand current situation in specific scenarios, valuable knowledge should be mined from both historical data and emerging new data. However, most existing algorithms take the historical data and the emerging data as a whole and periodically repeat to analyze all of them, which results in heavy computation overhead. It is also challenging to accurately discover new knowledge in time, because the emerging data are usually small compared to the historical data. To address these challenges, we propose a novel knowledge discovery algorithm based on double evolving frequent pattern trees that can trace the dynamically evolving data by an incremental sliding window. One tree is used to record frequent patterns from the historical data, and the other one records incremental frequent items. The structures of the double frequent pattern trees and their relationships are updated periodically according to the emerging data and a sliding window. New frequent patterns are mined from the incremental data and new knowledge can be obtained from pattern changes. Evaluations show that this algorithm can discover new knowledge from evolving data with good performance and high accuracy.

An Evolutive Frequent Pattern Tree-based Incremental Knowledge Discovery Algorithm

Social data from online social networks is expanding rapidly as the number of users and articles posted increases, making public opinion analysis a greater challenge. Real-time topic detection is a key part of public opinion analysis. The complex data processing involved in traditional clustering and text categorization can lead to time delays in topic detection. In this paper we construct similar networks and detect topics from similar communities that reduces the processing overhead in obtaining real-time topics. The similar communities consist of users with high similarity between them. We collect public topics from the microposts of delegates selected from each similar community. Selecting delegates can reduce the processing time of large amounts of redundant data during topic detection. Obtaining public opinion keywords in real time allows organizations to respond to public opinion security incidents in real time. Experiments showed that our scheme can find public topics faster and more effectively than two traditional algorithms.

Topic detection based on similar networks

Ship plate recognition is challenging due to variations of plate locations and text types. This paper proposes an effcient Fully Convolutional Network based Plate Recognition approach FCNPR, which uses a CNN (Convolutional Neural Network) to locate ships, then detects plate text lines with the fully convolutional network (FCN). The recognition accuracy is improved with integrating the AIS (Automatic Identification System) information. The actual FCNPR deployment demonstrates that it can work reliably with a high accuracy for satisfying practical usages.

/pdf/fully-convolutional-network-based-ship-plate-recognition-3jzevkkirn.pdf

Xin Liu

Papers

POCA4SD: A Public Opinion Cellular Automata for Situation Deduction

Towards a High Speed Video Cloud Based on Batch Processing Integrated with Fast Processing

An Evolutive Frequent Pattern Tree-based Incremental Knowledge Discovery Algorithm

Topic detection based on similar networks

Fully Convolutional Network Based Ship Plate Recognition