Ensembles of relational classifiers

Relational classification aims at including relations among entities into the classification process, for example taking relations among documents such as common authors or citations into account. However, considering more than one relation can further improve classification accuracy. Here we introduce a new approach to make use of several relations as well as both, relations and local attributes for classification using ensemble methods. To accomplish this, we present a generic relational ensemble model that can use different relational and local classifiers as components. Furthermore, we discuss solutions for several problems concerning relational data such as heterogeneity, sparsity, and multiple relations. Especially the sparsity problem will be discussed in more detail. We introduce a new method called PRNMultiHop that tries to handle this problem. Furthermore we categorize relational methods in a systematic way. Finally, we provide empirical evidence, that our relational ensemble methods outperform existing relational classification methods, even rather complex models such as relational probability trees (RPTs), relational dependency networks (RDNs) and relational Bayesian classifiers (RBCs).     

SELECTING EFFECTIVE FEATURES AND RELATIONS FOR EFFICIENT MULTI‐RELATIONAL CLASSIFICATION

Feature selection is an essential data processing step to remove irrelevant and redundant attributes for shorter learning time, better accuracy, and better comprehensibility. A number of algorithms have been proposed in both data mining and machine learning areas. These algorithms are usually used in a single table environment, where data are stored in one relational table or one flat file. They are not suitable for a multi‐relational environment, where data are stored in multiple tables joined to one another by semantic relationships. To address this problem, in this article, we propose a novel approach called FARS to conduct both Feature And Relation Selection for efficient multi‐relational classification. Through this approach, we not only extend the traditional feature selection method to select relevant features from multi‐relations, but also develop a new method to reconstruct the multi‐relational database schema and eliminate irrelevant tables to improve classification performance further. The results of the experiments conducted on both real and synthetic databases show that FARS can effectively choose a small set of relevant features, thereby enhancing classification efficiency and prediction accuracy significantly.     

Efficient regularized least-squares algorithms for conditional ranking on relational data

In domains like bioinformatics, information retrieval and social network analysis, one can find learning tasks where the goal consists of inferring a ranking of objects, conditioned on a particular target object. We present a general kernel framework for learning conditional rankings from various types of relational data, where rankings can be conditioned on unseen data objects. We propose efficient algorithms for conditional ranking by optimizing squared regression and ranking loss functions. We show theoretically, that learning with the ranking loss is likely to generalize better than with the regression loss. Further, we prove that symmetry or reciprocity properties of relations can be efficiently enforced in the learned models. Experiments on synthetic and real-world data illustrate that the proposed methods deliver state-of-the-art performance in terms of predictive power and computational efficiency. Moreover, we also show empirically that incorporating symmetry or reciprocity properties can improve the generalization performance.     

A novel recommendation approach based on users’ weighted trust relations and the rating similarities

With the growing popularity of open social networks, approaches incorporating social relationships into recommender systems are gaining momentum, especially matrix factorization-based ones. The experiments in previous literatures indicate that social information is very effective in improving the performance of traditional recommendation algorithms. However, most of existing social recommendation methods only take one kind of social relations—trust information into consideration, which is far from satisfactory. Furthermore, most of the existing trust networks are binary, which results in the equal treatment to different users who are trusted by the same user in these methods. In this paper, based on matrix factorization methods, we propose a new approach to make recommendation with social information. Its novelty can be summarized as follows: (1) it shows how to add different weights on the social trust relationships among users based on the trustee’s competence and trustworthiness; (2) it incorporates the similarity relationships among users as a complement into the social trust relationships to enhance the computation of user’s neighborhood; (3) it can balance the influence of these two kinds of relationships based on user’s individuality adaptively. Experiments on Epinions and Ciao datasets demonstrate that our approach outperforms the state-of-the-art algorithms in terms of mean absolute error and root mean square error, in particular for the users who rated a few items.     

FlowTrust: trust inference with network flows

Web-based social networking is increasingly gaining popularity due to the rapid development of computer networking technologies. However, social networking applications still cannot obtain a wider acceptance by many users due to some unresolved issues, such as trust, security, and privacy. In social networks, trust is mainly studied whether a remote user behaves as expected by an interested user via other users, who are respectively named trustee, trustor, and recommenders. A trust graph consists of a trustor, a trustee, some recommenders, and the trust relationships between them. In this paper, we propose a novel FlowTrust approach to model a trust graph with network flows, and evaluate the maximum amount of trust that can flow through a trust graph using network flow theory. FlowTrust supports multi-dimensional trust. We use trust value and confidence level as two trust factors.We deduce four trust metrics from these two trust factors, which are maximum flow of trust value, maximum flow of confidence level, minimum cost of uncertainty with maximum flow of trust, and minimum cost of mistrust with maximum flow of confidence. We also propose three FlowTrust algorithms to normalize these four trust metrics. We compare our proposed FlowTrust approach with the existing RelTrust and CircuitTrust approaches. We show that all three approaches are comparable in terms of the inferred trust values. Therefore, FlowTrust is the best of the three since it also supports multi-dimensional trust.     

基于无监督集成聚类的开放关系抽取方法

开放关系抽取(Open Relation Extraction, OpenRE)旨在从开放域语料库中抽取关系事实。大多数OpenRE方法通常局限于无监督方法提取命名实体之间的关系模式,然后将语义等价的模式聚类成一个关系簇,但由于缺少监督信息且聚类精度较低,影响了最终的关系抽取效果。为了进一步提高聚类性能,该文提出一种无监督集成聚类框架(Unsupervised Ensemble Clustering,UEC),它将无监督集成学习与基于信息度量的多步聚类算法相结合自主创建高质量伪标签,并以此作为监督信息改进关系特征的学习,从而引导聚类过程,获得更好的标签质量,最后通过多次迭代聚类发现文本中的关系类型。在FewRel和NYT-FB数据集上的实验结果表明,该文方法优于其他主流的基线OpenRE模型,F₁值分别达到了65.2%和67.1%。     

A correlative denoising autoencoder to model social influence for top-N recommender system

In recent years,there are numerous works been proposed to leverage the techniques of deep learning to improve social-aware recommendation performance.In most cases,it requires a larger number of data to train a robust deep learning model,which contains a lot of parameters to fit training data.However,both data of user ratings and social networks are facing critical sparse problem,which makes it not easy to train a robust deep neural network model.Towards this problem,we propose a novel correlative denoising autoencoder(CoDAE)method by taking correlations between users with multiple roles into account to learn robust representations from sparse inputs of ratings and social networks for recommendation.We develop the CoDAE model by utilizing three separated autoencoders to learn user features with roles of rater,truster and trustee,respectively.Especially,on account of that each input unit of user vectors with roles of truster and trustee is corresponding to a particular user,we propose to utilize shared parameters to learn common information of the units that corresponding to same users.Moreover,we propose a related regularization term to learn correlations between user features that learnt by the three subnetworks of CoDAE model.We further conduct a series of experiments to evaluate the proposed method on two public datasets for Top-N recommendation task.The experimental results demonstrate that the proposed model outperforms state-of-the-art algorithms on rank-sensitive metrics of MAP and NDCG.     

An interoperable context sensitive model of trust

Although the notion of trust is widely used in secure information systems, very few works attempt to formally define it or reason about it. Moreover, in most works, trust is defined as a binary concept—either an entity is completely trusted or not at all. Absolute trust on an entity requires one to have complete knowledge about the entity. This is rarely the case in real-world applications. Not trusting an entity, on the other hand, prohibits all communications with the entity rendering it useless. In short, treating trust as a binary concept is not acceptable in practice. Consequently, a model is needed that incorporates the notion of different degrees of trust. We propose a model that allows us to formalize trust relationships. The trust relationship between a truster and a trustee is associated with a context and depends on the experience, knowledge, and recommendation that the truster has with respect to the trustee in the given context. We show how our model can measure trust and compare two trust relationships in a given context. Sometimes enough information is not available about a given context to evaluate trust. Towards this end we show how the relationships between different contexts can be captured using a context graph. Formalizing the relationships between contexts allows us to extrapolate values from related contexts to approximate the trust of an entity even when all the information needed to calculate the trust is not available. Finally, we show how the semantic mismatch that arises because of different sources using different context graphs can be resolved and the trust of information obtained from these different sources compared.

Clustering and aggregation of relational data with applications to image database categorization

In this paper, we introduce a new algorithm for clustering and aggregating relational data (CARD). We assume that data is available in a relational form, where we only have information about the degrees to which pairs of objects in the data set are related. Moreover, we assume that the relational information is represented by multiple dissimilarity matrices. These matrices could have been generated using different sensors, features, or mappings. CARD is designed to aggregate pairwise distances from multiple relational matrices, partition the data into clusters, and learn a relevance weight for each matrix in each cluster simultaneously. The cluster dependent relevance weights offer two advantages. First, they guide the clustering process to partition the data set into more meaningful clusters. Second, they can be used in subsequent steps of a learning system to improve its learning behavior. The performance of the proposed algorithm is illustrated by using it to categorize a collection of 500 color images. We represent the pairwise image dissimilarities by six different relational matrices that encode color, texture, and structure information.     

Fuzzy relations and fuzzy relational databases

This paper deals with the connections existing between fuzzy set theory and fuzzy relational databases. Our new result dealing with fuzzy relations is how to calculate the greatest lower bound (glb) of two similarity relations. Our main contributions in fuzzy relational databases are establishing from fuzzy set theory what a fuzzy relational database should be (the result is both surprising and elegant), and making fuzzy relational databases even more robust.Our work in fuzzy relations and in fuzzy databases had led us into other interesting problems—two of which we mention in this paper. The first is primarily mathematical, and the second provides yet another connection between fuzzy set theory and artificial intelligence. In understanding similarity relations in terms of other fuzzy relations and in making fuzzy databases more robust, we work with closure and interior operators; we present some important properties of these operators. In establishing the connection between fuzzy set theory and artificial intelligence, we show that an abstraction on a set is in fact a partition on the set; that is, an abstraction defines an equivalence relation on the underlying set.     

Optimizing updates of recursive XML views of relations

XML publishing has been an emerging technique for transforming (portions of) a relational database into an XML document, for example, to facilitate interoperability between heterogeneous applications. Such applications may update the XML document and the source relational database must be updated accordingly. In this paper, we consider such XML documents as (possibly) recursively defined XML views of relations. We propose new optimization techniques to efficiently support XML view updates specified via an XPATH expression with recursion and complex filters. The main novelties of our techniques are: (1) we propose a space-efficient relational encoding of recursive XML views; and (2) we push the bulk of update processing inside a relational database. Specifically, a compressed representation of the XML views is stored as extended shared-inlining relations. A space-efficient and updatable 2-hop index is used to optimize XPATH evaluation on XML views. Updates of the XML views are evaluated on these relations and index. View update translation is handled by a heuristic procedure inside a relational database, as opposed to previous middleware approaches. We present an experimental study to demonstrate the effectiveness of our proposed techniques.     

The Mind’s Eye on Personal Profiles: A Cognitive Perspective on Profile Elements that Inform Initial Trustworthiness Assessments and Social Awareness in Virtual Project Teams

Collaboration in virtual project teams heavily relies on interpersonal trust, for which perceived professional trustworthiness is an important determinant. In face to face teams colleagues form a first impression of each others trustworthiness based on signs and signals that are ‘naturally’ available. However, virtual project team members do not have the same opportunities to assess trustworthiness. This study provides insight in the information elements that virtual project team members value to assess professional trustworthiness in the initial phase of collaboration. The trustworthiness formed initially is highly influential on interpersonal trust formed during latter collaboration. We expect trustors in virtual teams to especially value information elements (= small containers for personal data stimulating the availability of specific information) that provide them with relevant cues of trust warranting properties of a trustee. We identified a list with fifteen information elements that were highly valued across trustors (n?=?226) to inform their trustworthiness assessments. We then analyzed explanations for preferences with the help of a theory-grounded coding scheme for perceived trustworthiness. Results show that respondents value those particular information elements that provide them with multiple cues (signaling multiple trust warranting properties) to assess the trustworthiness of a trustee. Information elements that provide unique cues (signaling for a specific trust warranting property) could not be identified. Insight in these information preferences can inform the design of artefacts, such as personal profile templates, to support acquaintanceships and social awareness especially in the initial phase of a virtual project team.     

A relational, indirect, meso-level approach to CSCL design in the next decade

This paper reviews some foundational issues that we believe will affect the progress of CSCL over the next ten years. In particular, we examine the terms technology, affordance, and infrastructure and we propose a relational approach to their use in CSCL. Following a consideration of networks, space, and trust as conditions of productive learning, we propose an indirect approach to design in CSCL. The work supporting this theoretical paper is based on the outcomes of two European research networks: E-QUEL, a network investigating e-quality in e-learning; and Kaleidoscope, a European Union Framework 6 Network of Excellence. In arguing for a relational understanding of affordance, infrastructure, and technology we also argue for a focus on what we describe as meso-level activity. Overall this paper does not aim to be comprehensive or summative in its review of the state of the art in CSCL, but rather to provide a view of the issues currently facing CSCL from a European perspective.     

A method for reduction of examples in relational learning

Feature selection methods often improve the performance of attribute-value learning. We explore whether also in relational learning, examples in the form of clauses can be reduced in size to speed up learning without affecting the learned hypothesis. To this end, we introduce the notion of safe reduction: a safely reduced example cannot be distinguished from the original example under the given hypothesis language bias. Next, we consider the particular, rather permissive bias of bounded treewidth clauses. We show that under this hypothesis bias, examples of arbitrary treewidth can be reduced efficiently. We evaluate our approach on four data sets with the popular system Aleph and the state-of-the-art relational learner nFOIL. On all four data sets we make learning faster in the case of nFOIL, achieving an order-of-magnitude speed up on one of the data sets, and more accurate in the case of Aleph.     

Context-aware trust network extraction in large-scale trust-oriented social networks

In recent years, social networking sites have been used as a means for a rich variety of activities, such as movie recommendations and product recommendations. In order to evaluate the trust between a truster (i.e., the source) and a trustee (i.e., the target) who have no direct interaction in Online Social Networks (OSNs), the trust network between them that contains important intermediate participants, the trust relations between the participants, and the social context, has an important influence on trust evaluation. Thus, to deliver a reasonable trust evaluation result, before performing any trust evaluation (i.e., trust transitivity), the contextual trust network from a given source to a given target needs to be first extracted from the social network, where constraints on social context should also be considered to guarantee the quality of the extracted networks. However, this problem has been proved to be NP-Complete. Towards solving this challenging problem, we first present a contextual trust-oriented social network structure which takes social contextual impact factors, including trust, social intimacy degree, community impact factor, preference similarity and residential location distance into account. These factors have significant influences on both social interactions between participants and trust evaluation. Then, we present a new concept QoTN (Quality of Trust Network) and propose a social context-aware trust network extraction model. Finally, we propose a Heuristic Social Context-Aware trust Network extraction algorithm (H-SCAN-K) by extending the K-Best-First Search (KBFS) method with several proposed optimization strategies. The experiments conducted on two real datasets illustrate that our proposed model and algorithm outperform the existing methods in both algorithm efficiency and the quality of the extracted trust networks.     

Semi-supervised online structure learning for composite event recognition

Online structure learning approaches, such as those stemming from statistical relational learning, enable the discovery of complex relations in noisy data streams. However, these methods assume the existence of fully-labelled training data, which is unrealistic for most real-world applications. We present a novel approach for completing the supervision of a semi-supervised structure learning task. We incorporate graph-cut minimisation, a technique that derives labels for unlabelled data, based on their distance to their labelled counterparts. In order to adapt graph-cut minimisation to first order logic, we employ a suitable structural distance for measuring the distance between sets of logical atoms. The labelling process is achieved online (single-pass) by means of a caching mechanism and the Hoeffding bound, a statistical tool to approximate globally-optimal decisions from locally-optimal ones. We evaluate our approach on the task of composite event recognition by using a benchmark dataset for human activity recognition, as well as a real dataset for maritime monitoring. The evaluation suggests that our approach can effectively complete the missing labels and eventually, improve the accuracy of the underlying structure learning system.

     

An enhanced e-commerce trust model for community based centralized systems

Conventional trust models cannot simultaneously fulfill three major requirements of community based centralized e-commerce systems having a large number of members and economic risk. First, a buyer dealing with a new seller should be able to consult other buyers’ experiences with that seller to make a better decision. Also, a seller with a good reputation should not abuse it by conducting fraudulent transactions. Finally, a buyer with good experience with a seller, even one with a bad reputation, should be able to continue to deal with that seller. This paper introduces a trust-scoring model (Enhanced E-Commerce Trust Model or E2CTM) to address the above requirements. In E2CTM, a trustor uses personal experience with a trustee and input from other trustors about that trustee. To explore the capability of E2CTM, an online electronic auction shopping system was developed to compare its performance with a conventional trust model used by most electronic shopping systems that uses overall gain and loss of all buyers in simulated auctions as evaluation criteria. The results show the advantage of E2CTM, specifically, in terms of overall performance. The improvement is more significant when there is low percentage of fraudulent transactions in the system.     

A bias/variance decomposition for models using collective inference

Bias/variance analysis is a useful tool for investigating the performance of machine learning algorithms. Conventional analysis decomposes loss into errors due to aspects of the learning process, but in relational domains, the inference process used for prediction introduces an additional source of error. Collective inference techniques introduce additional error, both through the use of approximate inference algorithms and through variation in the availability of test-set information. To date, the impact of inference error on model performance has not been investigated. We propose a new bias/variance framework that decomposes loss into errors due to both the learning and inference processes. We evaluate the performance of three relational models on both synthetic and real-world datasets and show that (1) inference can be a significant source of error, and (2) the models exhibit different types of errors as data characteristics are varied.     

Correcting evaluation bias of relational classifiers with network cross validation

Recently, a number of modeling techniques have been developed for data mining and machine learning in relational and network domains where the instances are not independent and identically distributed (i.i.d.). These methods specifically exploit the statistical dependencies among instances in order to improve classification accuracy. However, there has been little focus on how these same dependencies affect our ability to draw accurate conclusions about the performance of the models. More specifically, the complex link structure and attribute dependencies in relational data violate the assumptions of many conventional statistical tests and make it difficult to use these tests to assess the models in an unbiased manner. In this work, we examine the task of within-network classification and the question of whether two algorithms will learn models that will result in significantly different levels of performance. We show that the commonly used form of evaluation (paired t-test on overlapping network samples) can result in an unacceptable level of Type I error. Furthermore, we show that Type I error increases as (1) the correlation among instances increases and (2) the size of the evaluation set increases (i.e., the proportion of labeled nodes in the network decreases). We propose a method for network cross-validation that combined with paired t-tests produces more acceptable levels of Type I error while still providing reasonable levels of statistical power (i.e., 1−Type II error).     

Professional confidence in the roles as ambulance and medical incident commander

The aim was to investigate professional confidence in the roles of ambulance and medical incident commander (AIC and MIC), and how it influences achievement of performance indicators at an incident site. A web survey based on theoretical constructs (e.g., social identity, efficacy, accountability) and questions about prehospital emergency care connected to the roles were used (n = 426 Swedish ambulance nurses and emergency medical technicians). The results showed that social identity, independence and occupation were moderators for professional confidence. Organizational support, relational trust and independence were moderators for achieving performance indicators. Strengthening group identification and independence as MIC and independence and support for women as AIC together with a stronger organizational support can increase professional confidence and improve performance.