In this paper, we present a novel method Coarse- and Fine-grained Attention Network (CFANet) for generating high-quality crowd density maps people count estimation by incorporating attention to better focus on the area. We devise from-coarse-to-fine progressive mechanism integrating Crowd Region Recognizer (CRR) Density Level Estimator (DLE) branch, which can suppress influence of irrelevant background assign weights according levels, because accurate fine-grained directly is normally...

Collaborative filtering (CF) recommenders based on User-Item rating matrix as explicitly obtained from end users have recently appeared promising in recommender systems. However, is not always available or very sparse some web applications, which has critical impact to the application of CF recommenders. In this article we aim enhance online system by fusing virtual ratings derived user reviews. Specifically, taking into account Chinese reviews' characteristics, propose fuse self-supervised...

The increasing availabilities of GPS-enabled devices have given rise to the location-based social networking services (LBSN), in which users can record their travel experiences with GPS trajectories and share these among each other on Web communities. Usually, far denser points than necessary scenarios GPS-trajectory-sharing. Meanwhile, redundant will decrease performance LBSN systems even cause browser crashed. Existing line simplification algorithms only focus maintaining shape information...

Applications of various data analytics technologies to security and criminal investigation during the past three decades have demonstrated inception, growth, maturation analytics. We first identify five cutting‐edge mining such as link analysis, intelligent agents, text mining, neural networks, machine learning. Then, we explore their recent applications domain, discuss challenges arising from these innovative applications. also extend our study big which provides some state‐of‐the‐art...

In this paper, we propose a novel approach to classify short texts by combining both their lexical and semantic features. We present an improved measurement method for feature selection furthermore obtain the features with background knowledge repository which covers target category domains. The combination of is achieved mapping words topics different weights. way, dimensionality space reduced number topics. here use Wikipedia as employ Support Vector Machine (SVM) classifier. experiment...

Although much research is devoted to the analysis and prediction of individuals' behavior in social networks, very few studies analyze firms' performance with respect business networks. Empowered by recent on automated mining this article illustrates design a novel network-based model called energy cascading (ECM) for predicting directional stock price movements related firms. More specifically, proposed predictive analytics considers both influential relationships Twitter sentiments infer...

Topic models have been prevailing for many years on discovering latent semantics while modeling long documents. However, short texts they generally suffer from data sparsity because of extremely limited word co-occurrences; thus tend to yield repetitive or trivial topics with low quality. In this paper, address issue, we propose a novel neural topic model in the framework autoencoding new distribution quantization approach generating peakier distributions that are more appropriate texts....

Dirty data commonly exist. Simply discarding a large number of inaccurate points (as noises) could greatly affect clustering results. We argue that dirty can be repaired and utilized as strong supports in clustering. To this end, we study novel problem repairing over at the same time. Referring to minimum change principle repairing, objective is find modification such amount enhance show formulated an integer linear programming (ILP) problem. Efficient approximation then devised by (LP)...

The goal of entity matching is to find the corresponding records representing same from different data sources. At present, in mainstream methods, rule-based methods need tremendous domain knowledge. Machine-learning-based or deep-learning-based a large number labeled samples build model, which difficult achieve some applications. In addition, learning-based are more likely overfit, so quality requirements training very high. this paper, we present an active learning method for tasks. This...

Community Question Answering (CQA) services provide an open platform for people to share their knowledge and have attracted great attention its rapidly increasing popularity. As more is shared in CQA, how use the repository solving new questions has become a crucial problem. In this paper, we tackle problem by finding experts from question answering history first then recommending appropriate answer questions. We develop Topic-level Expert Learning (TEL) model find on topic level CQA. Our...

We study the problem of social recommendation incorporating topic mining and trust analysis. Different from other works related to recommendation, we merge analysis techniques into recommender systems for finding topics tags items estimating topic-specific trust. propose a probabilistic matrix factorization (TTMF) algorithm try enhance accuracy by utilizing estimated relations. Moreover, TTMF is also convenient solve item cold start inferring feature (topic) new their tags. Experiments are...

In Human Activity Recognition (HAR), supervised Machine Learning methods are predominantly used, making availability of datasets a major issue for research in the field. particular, majority available collected under controlled conditions. Consequently, models trained similar circumstances, generally exhibit significant decrease recognition accuracy when they moved to final deployment wild, within unconstrained settings. This paper presents new dataset HAR, free-living and 10 subjects were...

Since effective semantic representations are utilized in many practical applications, inferring discriminative and coherent latent topics from short texts is a critical basic task. Traditional topic models like Probabilistic Latent Semantic Analysis (PLSA) Dirichlet Allocation (LDA) behave not well on due to data sparsity problem. One novel model called Biterm Topic Model (BTM) which unordered word-pairs (i.e., biterms) whole corpus was proposed solve this However, both the performance...

Big data analytics have been proposed as a disruptive technology that will reshape security and law enforcement agencies, which is domain relies on to achieve criminal network insights for better decision-making. Rooted in current literature, we review the landscape of analysis through big analytic framework this paper. We identify sources, platforms, tools, applications related analysis. then present some real-world examples, research challenges, future directions applying domain.