added 3 research items
This paper presents a novel matrix factorization (MF) model, called FeatureMF, which takes into account item features and thus addresses the cold-start item and data sparsity problems of collaborative filtering (CF). More specifically, the model extends item latent vectors with item representation learned from metadata. Experiments conducted on a public dataset with two testing views confirm that FeatureMF achieves better recommendation performance than some of the popular state-ofthe-art MF models.
This paper presents a novel matrix factorization (MF) recommendation model, FeatureMF, which extends item latent vectors with item representation learned from metadata. By taking into account item features, the model addresses the coldstart item problem and data-sparsity problem of collaborative filtering (CF). Extensive experiments conducted on a public dataset with two testing views confirm that FeatureMF achieves better prediction accuracy than some of the popular state-of-theart MF-based recommendation models.
This paper proposes an improvement to item recommendation systems based on collaborative filtering (CF) with implicit feedback data. Combined with the Bayesian Personalized Ranking (BPR) optimization approach, recommended for implicit-only feedback contexts, CF has been shown to be effective in generating accurate recommendations. The method, based on the assumption that a user prefers a consumed item to an unconsumed item, aims to maximize the difference of predicted scores between these items for each user. In most of the existing CF recommendation methods, all items are assigned the same weight, which of course is not the case in reality. In this paper, a new improved matrix factorization (MF) approach is proposed where the weights of items are allowed to vary and be reflective of items' importance or their desirability to a user. The scheme integrates these item weights as appropriate and utilizes a dynamic learning model where learning is driven by BPR. The performance of the proposed method is tested against the traditional MF. Tests confirm that better accuracy can be indeed achieved by the proposed method.
Recommendation systems employed on the Internet aim to serve users by recommending items which will likely be of interest to them. The recommendation problem could be cast as either a rating estimation problem which aims to predict as accurately as possible for a user the rating values of items which are yet unrated by that user, or as a ranking problem which aims to find the top-k ranked items that would be of most interest to a user, which s/he has not ranked yet. In contexts where explicit item ratings of other users may not be available, the ranking prediction could be more important than the rating prediction. Most of the existing ranking-based prediction approaches consider items as having equal weights which is not always the case. Different weights of items could be regarded as a reflection of items' importance, or desirability, to users. In this paper, we propose to integrate variable item weights with a ranking-based matrix factorization model, where learning is driven by Bayesian Personalized Ranking (BPR). Two ranking-based models utilizing different-weight learning methods are proposed and the performance of both models is confirmed as being better than the standard BPR method.
Exploiting additional item meta-data is proposed in this paper for solving data sparsity and cold start problems found in item-based collaborative filtering (CF) techniques, which are employed in recommendation systems. Additional item meta-data provides the foundation for generating a heterogeneous information network (HIN). The proposed approach is to enrich the item-based CF with diverse types of relationships existing between items in the HIN, to overcome the sparsity issue from implicit user feedback. Bayesian personalized ranking optimization technique is used for estimation and its performance is evaluated by comparing the results with the traditional item-based CF. The experimental tests prove that the proposed approach achieves better accuracy.
With the development of ubiquitous computing, recommendation systems have become essential tools in assisting users in discovering services they would find interesting. This process is highly dynamic with an increasing number of services, distributed over networks, bringing the problems of cold start and sparsity for service recommendation to a new level. To alleviate these problems, this paper proposes a hybrid service recommendation prototype utilizing user and item side information, which naturally constitute a heterogeneous information network (HIN) for use in the emerging ubiquitous consumer wireless world (UCWW) wireless communication environment that offers a consumer-centric and network-independent service operation model and allows the accomplishment of a broad range of smart-city scenarios, aiming at providing consumers with the “best” service instances that match their dynamic, contextualized, and personalized requirements and expectations. A layered architecture for the proposed prototype is described. Two recommendation models defined at both global and personalized level are proposed, with model learning based on the Bayesian Personalized Ranking (BPR). A subset of the Yelp dataset is utilized to simulate UCWW data and evaluate the proposed models. Empirical studies show that the proposed recommendation models outperform several widely deployed recommendation approaches.
With the rapid growth of the Web, recommender systems have become essential tools to assist users to find high-quality personalized recommendations from massive information resources. Content-based filtering (CB) and collaborative filtering (CF) are the two most popular and widely used recommendation approaches. In this paper, we focus on ways of taking advantage of both approaches based only on user-item rating data. Motivated by the user profiling technique used in content-based recommendation, we propose to merge user profiles, learnt from the items viewed by the users, as a new latent variable in the latent factor model, which is one of the most popular CF-based approaches, thereby generating more accurate recommendation models. The performance of the proposed models is tested against several widely-deployed state-of-the-art recommendation methods. Experimental results, based on two popular datasets, confirm that better accuracy can be indeed achieved.
Context-aware recommendation systems make recommendations by adapting to user's specific situation, and thus by exploring both the user preferences and the environment. In this paper, the design of a context-aware service recommendation framework utilising semantic knowledge in the Ubiquitous Consumer Wireless World (UCWW) is outlined. The main objective of the framework is to point to users the 'best' service instances that match their dynamic, contextualised and personalised requirements and expectations, thereby aligning to the always best connected and best served (ABC&S) paradigm.
A wireless solution for context- and service-awareness in mobile communications is the theme of this paper. Respecting mobile users’ desire for minimal intrusion of unsolicited advertisements, here we show how the novel push-advertisement technology and medium of ‘wireless billboard channels’ (WBCs) could be employed by service providers to broadcast advertisements of their wireless services to mobile terminals. While the word ‘billboard’ here seems to hint at in-your-face intrusiveness, in fact the service functions in the background in a near-transparent un-intrusive manner to the user. When combined with user-driven, dynamic, smart user-profile functionality, inclusive of smart optimisation of current user’s ‘always best connected and best served’ (ABC&S) policies, the system has the potential to provide an effective, pro-active, wireless-based, context-aware and service-aware infrastructure. The exposition here includes detailed descriptions of the WBC concept, its associated advertisement, discovery and association (ADA) functionality, full technical details of the WBC advertisement service description techniques, formats and attributes, and operational aspects, such as the WBC bit stream structure. It also includes discussion of algorithmic approaches towards optimising smart user profile functionality on mobile terminal which will drive the ABC&S decision-making in ways matched especially to the user needs, in particular those schemes utilising Personalised Information Retrieval (PIR) systems.
This paper describes the design and development of a novel cloud-based system for increased service contextualization in future wireless networks. The principal objective is the support of mobile users (consumers) in a Ubiquitous Consumer Wireless World (UCWW) seeking to choose and select the ‘best’ service instance in a UCWW environment matched to their dynamic contextualized and personalized service delivery requirements and expectations, thereby increasing user freedom in where, when and how they access desired services, and increasing user-driven networking. The design challenges to create such a cloud-based system with an ever-enhanced capacity to be attuned to a user-client’s dynamic contexts, and do this for all its users, are addressed, and software infrastructural design solutions suggested. The cloud idea proposed here is one which should yield efficiencies and saving for consumers, operate as an additional ‘behind-the-scenes’ decision support subsystem to make smart decisions based on mining of the most up-to-date data stored in the cloud repositories related to service contexts and personalized profiles. Rather than the use of known efficient heuristic methods employed with large and complex data structures, together with associated algorithms solving the combinatorial optimization problems, an alternative method, proposed here for making predictions, is to discover patterns in the behaviour of the individual clientconsumer, to bring into play, in the decision process, patterns and trends of other client-consumers seeking the same or similar services, and also the constant update of the user’s wireless environment context through information garnered from other sources, such as wireless access service provider updates, teleservice provider updates, and data sensed by the sensors in the environment. Indentifying and addressing the need as directly as this is a novel approach towards providing context-aware personalized services. It is particularly novel, and desirable, in the UCWW context. Hence, this consumer supportive smart repository solution may appropriately be called a UCWW cloud. The paper sets out an infrastructural design of this cloud, ordered within a conceptual UCWW software architecture, together with its various elements, e.g., decision support subsystem and mobile network environment elements of personalized information retrieval (PIR).