Implicit Recommendation Research Articles

Targeted Advertising as Implicit Recommendation: Strategic Mistargeting and Personal Data Opt-out

Targeted Advertising as Implicit Recommendation: Strategic Mistargeting and Personal Data Opt-out

GPR-OPT: A Practical Gaussian optimization criterion for implicit recommender systems

Implicit recommendation refers to the users’ feedback on items derived from their interactions with items, i.e., clicks, and purchases. The methods in the implicit recommendation scenario usually regard all the adopted items as their favorites and indiscriminately assign a uniform confidence weight of their preference toward all the adopted items. In practice, however, a user’s preferences toward different items vary a lot. Treating them equally in existing implicit feedback recommender systems may limit the capacity of learning algorithms. To address this problem, we propose a novel Gaussian Personalized Recommendation OPTimization criterion (GPR-OPT), and our aim is to make the unknown preference confidence of users toward their adopted items in implicit feedback recommendation be learnable, so as to improve the accuracy of implicit recommender systems. In particular, we assume the user’s interests in items follow Gaussian distributions. By maximizing the posterior probability of items derived from the Gaussian distribution of user features, GPR-OPT is able to self-adaptively learn the confidence of users’ preferences from the implicit user–item interactions. We conduct extensive experiments on three real-world datasets, i.e., Movielens 1M, Amazon Book, and Yelp, which show an average of 11.64% improvements over different kinds of collaborative filtering algorithms. GPR-OPT is a generic optimization criterion and can be easily integrated into most existing collaborative filtering recommendation models, leading to a great impact on implicit recommender systems.

Criterion-based Heterogeneous Collaborative Filtering for Multi-behavior Implicit Recommendation

Recent years have witnessed the explosive growth of interaction behaviors in multimedia information systems, where multi-behavior recommender systems have received increasing attention by leveraging data from various auxiliary behaviors such as tip and collect. Among various multi-behavior recommendation methods, non-sampling methods have shown superiority over negative sampling methods. However, two observations are usually ignored in existing state-of-the-art non-sampling methods based on binary regression: (1) users have different preference strengths for different items, so they cannot be measured simply by binary implicit data; (2) the dependency across multiple behaviors varies for different users and items. To tackle the above issue, we propose a novel non-sampling learning framework named C riterion-guided H eterogeneous C ollaborative F iltering (CHCF). CHCF introduces both upper and lower thresholds to indicate selection criteria, which will guide user preference learning. Besides, CHCF integrates criterion learning and user preference learning into a unified framework, which can be trained jointly for the interaction prediction of the target behavior. We further theoretically demonstrate that the optimization of Collaborative Metric Learning can be approximately achieved by the CHCF learning framework in a non-sampling form effectively. Extensive experiments on three real-world datasets show the effectiveness of CHCF in heterogeneous scenarios.

Revisiting Negative Sampling vs. Non-sampling in Implicit Recommendation

Recommendation systems play an important role in alleviating the information overload issue. Generally, a recommendation model is trained to discern between positive (liked) and negative (disliked) instances for each user. However, under the open-world assumption, there are only positive instances but no negative instances from users’ implicit feedback, which poses the imbalanced learning challenge of lacking negative samples. To address this, two types of learning strategies have been proposed before, the negative sampling strategy and non-sampling strategy. The first strategy samples negative instances from missing data (i.e., unlabeled data), while the non-sampling strategy regards all the missing data as negative. Although learning strategies are known to be essential for algorithm performance, the in-depth comparison of negative sampling and non-sampling has not been sufficiently explored by far. To bridge this gap, we systematically analyze the role of negative sampling and non-sampling for implicit recommendation in this work. Specifically, we first theoretically revisit the objection of negative sampling and non-sampling. Then, with a careful setup of various representative recommendation methods, we explore the performance of negative sampling and non-sampling in different scenarios. Our results empirically show that although negative sampling has been widely applied to recent recommendation models, it is non-trivial for uniform sampling methods to show comparable performance to non-sampling learning methods. Finally, we discuss the scalability and complexity of negative sampling and non-sampling and present some open problems and future research topics that are worth being further explored.

Экономика «своих» и «чужих»: роль этической системы в хозяйственном развитии

This study examines how the influence of collectivist values on economic prosperity is contingent upon the ethical systems that establish a foundation for cooperation within a society. Cultural studies suggest that individualistic cultures tend to develop a system of generalized morality, which promotes ethical and coope­rative behavior among abstract individuals in a society. In contrast, collectivist societies often adopt an ethi­cal system characterized by limited morality, confining moral conduct to family and relatives while excluding other groups. Limited morality has been frequently attributed to poor economic performance in collecti­vist cultures. Consequently, a prevailing implicit recommendation in this body of literature is to adopt a universal moral system across all countries. Modern theories in anthropology and cultural psychology suggest that the distinction in behavioral norms between “in-group” and “out-group” is a durable cultural characteristic of collectivist countries. Hence, striving for a moral system based on generalized morality may not yield signifi­cant economic improvements. Therefore, we propose a hypothesis that an effective direction for state policy deve­lopment lies in expanding the concept of kinship beyond the confines of the family to encompass larger groups such as communities or companies, thereby fostering cooperation within these expanded units. To test this hypothesis, an econometric model has been developed, enabling the estimation of the differentiated impact of ethi­cal system types on income in individualistic and collectivist countries. To assess the level of collecti­vism on a national scale, we develop a novel kinship intensity index utilizing G. Murdock’s Ethnographic Atlas.

Halal Standard: Issues In Hotel Industry

The hotel plays an important role in the growing halal tourism industry. The Success of tourist visits often depends on how acceptable the hotel is, where they are staying with Muslim norms and facilities to conduct ibadah (worship) such as Sala (praying). Most hotels must take a holistic approach to obtain the Halal certification for their service operations practices. However, the services sector, such as tourism, has yet to tap into this lucrative market event though halal travel is gaining popularity globally. The research aims to identify the challenges for Halal hotel development in Malaysia and recognize the factors that meet Halal standards. Data collection was assembled through content analysis, review of journals, and articles by previous researchers. The implicit recommendation is that the industry builds division and integrity in ensuring Halal levels through Halal Insiders to ensure Halal compliance, which impacts organizational performance. The results of this paper will also indicate the importance for hoteliers to have the knowledge and prepare for the prospects and challenges they might face before developing Shariah-Compliant Hotel in Malaysia. Thus, extensive research is required in this area.

A probabilistic perspective on nearest neighbor for implicit recommendation

Over the past years, the recommender systems community invented several novel approaches that reached better and better prediction accuracy. Sequential recommendation, such as music recommendation, has seen large improvements from neural network-based models such as recurrent neural networks or transformers. When no sequential information is available or not relevant, such as for book, movie, or product recommendation, however, the classic k-nearest neighbor algorithm appears to remain competitive, even when compared to much more sophisticated methods. In this paper, we attempt to explain the inner workings of the nearest neighbor using probabilistic tools, treating similarity as conditional probability and presenting a novel model for explaining and removing popularity bias. First, we provide a probabilistic formulation of similarity and the classic prediction formula. Second, by modeling user behavior as a combination of personal preference and global influence, we are able to explain the presence of popularity bias in the predictions. Finally, we utilize Bayesian inference to construct a theoretically grounded variant of the widely used inverse frequency scaling, which we use to mitigate the effect of popularity bias in the predictions. By replacing the formerly ad hoc choices of nearest neighbor with probabilistically founded counterparts, we are able to improve prediction accuracy over a variety of data sets and gain an increased understanding of the theory behind the method.

Targeted Advertising as an Implicit Recommendation and Personal Data Opt-Out

Targeted Advertising as an Implicit Recommendation and Personal Data Opt-Out

Collaborative filtering grounded on knowledge graphs

Matrix Factorization (MF) is a widely used collaborative filtering technique for effectively modeling a user-item interaction in recommender system. Despite the successful application of MF and its variants, the method proves to be effective only in situations where there is an abundance of user-item interactions. However, user-item interaction data are usually sparse, limiting the effectiveness of the method. In addressing this problem, recent methods have proposed to use knowledge graphs (KGs) as additional information to complement the sparse user-item interaction data. This has proved challenging given the complexity of the KG structure. In this paper, we propose a collaborative filtering method that takes advantage of knowledge graphs. More specifically, the embedding of a user and item are both grounded on the item’s attributes in the knowledge graph, and are aggregated with generic user and item representations modeled by MF for implicit recommendation. Our model has demonstrated to outperform the recent state-of-the-art method KGCN [18] in very sparse settings, showing an effective integration of KGs in recommender systems.

Causal neural fuzzy inference modeling of missing data in implicit recommendation system

Implicit feedback data has many interaction form, such as clicking, collection and play count. These interactions can be tracked automatically, which can alleviate the problem of data sparsity caused by the difficulty of obtaining explicit ratings. Therefore, implicit recommendation attracts many scholars’ attentions. However, the uncertainty of the implicit feedback data takes a great challenge to the recommendation. Especially for the negative feedback (missing data), we could not be sure whether the users dislike or just have not seen the items. If we regard them as negative feedback, it may lead to bias of predictions. In this paper, we propose Causal Neural Fuzzy Inference (CNFI) algorithm to model the missing data in implicit recommendation. First, we give the mixed recommendation model based on CNFI. Specially, we analyze the impact factors of users exposure to items from causal perspective. Technically, fuzzy set theory is used for representing the impact factors. Causal neural fuzzy inference network is applied to learn the weights of fuzzy rules based on the fuzzy representation of impact factors. Finally, CNFI and Matrix Factorization (MF) perform joint learning to make recommendations. Extensive experiments on three realistic datasets verify the effectiveness and advancement of the proposed algorithm.

Tubes and bubbles topological confinement of YouTube recommendations.

The role of recommendation algorithms in online user confinement is at the heart of a fast-growing literature. Recent empirical studies generally suggest that filter bubbles may principally be observed in the case of explicit recommendation (based on user-declared preferences) rather than implicit recommendation (based on user activity). We focus on YouTube which has become a major online content provider but where confinement has until now been little-studied in a systematic manner. We aim to contribute to the above literature by showing whether recommendation on YouTube exhibits phenomena typical of filter bubbles, tending to lower the diversity of consumed content. Starting from a diverse number of seed videos, we first describe the properties of the sets of suggested videos in order to design a sound exploration protocol able to capture latent recommendation graphs recursively induced by these suggestions. These graphs form the background of potential user navigations along non-personalized recommendations. From there, be it in topological, topical or temporal terms, we show that the landscape of what we call mean-field YouTube recommendations is often prone to confinement dynamics. Moreover, the most confined recommendation graphs i.e., potential bubbles, seem to be organized around sets of videos that garner the highest audience and thus plausibly viewing time.

Fast Adaptively Weighted Matrix Factorization for Recommendation with Implicit Feedback

Recommendation from implicit feedback is a highly challenging task due to the lack of the reliable observed negative data. A popular and effective approach for implicit recommendation is to treat unobserved data as negative but downweight their confidence. Naturally, how to assign confidence weights and how to handle the large number of the unobserved data are two key problems for implicit recommendation models. However, existing methods either pursuit fast learning by manually assigning simple confidence weights, which lacks flexibility and may create empirical bias in evaluating user's preference; or adaptively infer personalized confidence weights but suffer from low efficiency.To achieve both adaptive weights assignment and efficient model learning, we propose a fast adaptively weighted matrix factorization (FAWMF) based on variational auto-encoder. The personalized data confidence weights are adaptively assigned with a parameterized neural network (function) and the network can be inferred from the data. Further, to support fast and stable learning of FAWMF, a new specific batch-based learning algorithm fBGD has been developed, which trains on all feedback data but its complexity is linear to the number of observed data. Extensive experiments on real-world datasets demonstrate the superiority of the proposed FAWMF and its learning algorithm fBGD.

Multimedia recommendation using Word2Vec-based social relationship mining

This study proposes the multimedia recommendation method using Word2Vec-based social relationship mining. This is to analyze users with a similar tendency on the basis of the keywords related to multimedia content and sentiment words of comments, to build a trust relationship, and to recommend multimedia. In order to solve the problem of data sparsity, metadata of multimedia content are collected and then are clustered by genre. User’s evaluate a preference for multimedia content. With the use of evaluation data, the attributes preferred by users are predicted. In terms of propensities, the sentiment words in users comments are classified by SVM on the basis of sentiment dictionary. The classified sentiment words are presented in vector with the use of Word2Vec. In terms of the vector of sentiment words, the dynamic relationship between users of words in the same preference by the similarity using the distance scale. It helps to build a trust relationship between users with preferences that can change with a lapse of time. Accordingly, multimedia content are recommended to users with a similar tendency. In terms of performance evaluation, F-measure is compared with the uses of precision and recall for a recommendation. As the result of evaluation, the social relationship mining method is evaluated to be better than explicit and implicit recommendation methods. With the proposed method, it is possible to search with metadata of content and make a intelligent recommendation explicitly and implicitly according to user’s tendency.

Blacklisting Recommendation System: Using Spatio-Temporal Patterns to Predict Future Attacks

In this paper, we study the problem of forecasting attack sources based on past attack logs from several contributors. We formulate this problem as an implicit recommendation system, and we propose a multi-level prediction model to solve it. Our model evaluates and combines various factors, namely: (i) attacker-victim history using time-series, (ii) attackers and/or victims interactions using neighborhood models and (iii) global patterns using singular value decomposition. We evaluate our combined method, referred to as Blacklisting Recommendation System (or BRS), on one month of logs from Dshield, and we demonstrate that it improves significantly the prediction rate over state-of-the-art methods as well as the robustness against poisoning attacks. Along the way, we analyze the Dshield dataset, and we reveal dominant patterns of malicious traffic.

Taskforce on Innovative International Financing for Health Systems: what next?

The high-level Taskforce on Innovative International Financing for Health Systems was set up in 2008 and chaired by Gordon Brown, Prime Minister of the United Kingdom of Great Britain and Northern Ireland, and Robert Zoellick, President of The World Bank. Its aim was to identify innovative and additional sources of funding for health systems strengthening in the 49 lowest-income countries of the world. The taskforce delivered its final recommendations in September 2009 together with two detailed working group papers.1,2 Here we summarize the main outputs and recommendations of the taskforce according to three areas: (i) costing the financing gap; (ii) new and innovative sources of finance; and (iii) making development assistance for health work better (Box 1). We then examine their limitations and propose further actions for the international health community. Box 1 Summary of recommendations by the high-level Taskforce on Innovative International Financing for Health Systems3 New sources of finance: Extend the mandatory solidarity levy on airline tickets to more countries (currently in place in several countries and used primarily to finance paediatric AIDS treatment through UNITAID); Explore the viability of levies on tobacco and currency transactions; Encourage voluntary private giving through: (i) voluntary levies on the purchase of airline tickets and mobile phone minutes (expected to raise US$ 3.2 billion by 2015); and (ii) a scheme called a “de-tax” which would earmark a share of value added tax receipts when participating businesses agree to add a share of their profits (estimated potential of US$ 220 million in 2010); Secure more private investment in health systems through establishing capital/risk mitigation fund(s). Out-sourcing to non-government providers and encouraging greater use of advanced market commitments, such as for vaccine purchases, were also mentioned as ways of securing investment from private sector actors. Making development assistance for health work better: More frontloading (i.e. concentrating payments at the beginning of an agreement) and predictability of aid, possibly by expanding the mandate of the International Financing Facility for Immunization. Expand the use of results-based “buy-down” (use of grant funding to reduce the cost of loans when specific performance targets are met) funding and more performance-based donor funding for the health sector. Establish a common health systems funding platform for the Global Fund to Fight AIDS, Tuberculosis and Malaria, the GAVI Alliance and The World Bank. Undertake a review of technical assistance, in view of evidence that it consumes a large proportion of aid and appears to be poor value for money. Gaps and challenges Costing Three issues stand out. The first relates to the models used for calculating the costs of scaling-up essential health services, including the assumptions on what is required to achieve that scale-up.1 The different models used by the taskforce did not just produce alternative costings, but also reflected different approaches to health systems strengthening as well as different levels of ambition. The World Health Organization (WHO)’s normative approach, for example, was bolder and advocated the simultaneous scaling-up of facility and community-based services, while The World Bank and the United Nations Children’s Fund (UNICEF) were less ambitious and advocated expanding low-cost, community-based services before undertaking any strengthening or expansion of facility-based services. In addition to the confusion of having different costing models, the taskforce reveals fundamental differences in opinion about the minimum requirements to strengthen health systems and the best way to expand coverage of essential health services. The second issue is that the individual country costings used to produce an aggregated “price tag” for all low-income countries are unavailable. And yet a full and proper discussion about the best way to fund and scale-up essential health services can ultimately only be conducted at the country level. In addition, the costings generated for health systems inputs such as “governance” are novel and need further empirical testing. A disaggregation of the data by country is therefore a vital next step. Third, an implicit recommendation of the taskforce is that a significant proportion of funding should come from private expenditure, in spite of the need to reduce the burden of health expenditure on poor households. This suggests that the required future funding from governments and donors has been underestimated.

Multiple Intelligences: Its Tensions and Possibilities

This article explores the tensions between Howard Gardner's theory of multiple intelligences and current educational policies emphasizing standardized and predictable outcomes. The article situates Gardner's theory within the historical interests among psychometricians in identifying those core processes that constitute human intelligence. Gardner's theory of multiple intelligences provides a significant contrast to the models of mind that have traditionally been used to understand how people think and make intelligent choices. The pursuit of a single G factor is contrasted with an array of specific intelligences in Gardner's conception. The implications of Gardner's view for education pertain to the cultivation of the various ways in which humans reflect intelligently and the implicit recommendation that individual proclivities, interests, and intelligences be cultivated. Such an approach to schooling would yield differences among the outcomes for children whose intelligences differed. It is this orientation to the aims of education that conflicts dramatically with a standards-driven approach to school improvement.

Multiple Intelligences: Its Tensions and Possibilities

This article explores the tensions between Howard Gardner's theory of multiple intelligences and current educational policies emphasizing standardized and predictable outcomes. The article situates Gardner's theory within the historical interests among psychometricians in identifying those core processes that constitute human intelligence. Gardner's theory of multiple intelligences provides a significant contrast to the models of mind that have traditionally been used to understand how people think and make intelligent choices. The pursuit of a single G factor is contrasted with an array of specific intelligences in Gardner's conception. The implications of Gardner's view for education pertain to the cultivation of the various ways in which humans reflect intelligently and the implicit recommendation that individual proclivities, interests, and intelligences be cultivated. Such an approach to schooling would yield differences among the outcomes for children whose intelligences differed. It is this orientation to the aims of education that conflicts dramatically with a standards-driven approach to school improvement.

X-ray Films of the Skull In Trauma Cases

To the Editor.— implicit recommendation given to the widespread use of x-ray films of the skull in all cases of head trauma, which was the subject ofLaw and Medicinein the Oct 30,1972, issue ofThe Journal, is contrary to the prudent practice of medicine. In fact, a previousLaw and Medicinearticle (Feb 28,1972) was a good deal more realistic on this same subject. In an article by Bell and Loop entitled The Utility and Futility of Radiographic Skull Examinations for Trauma, 1 it was determined that if skull examinations for trauma were reserved only for a group of patients demonstrating certain clinical findings and omitted in all other patients with trauma, the yearly saving in health care costs would be $15 million, with no adverse effect on the patient's medical care. In another article on the subject, Roberts and Shopfner 2 point out in the analysis