No description available.

Identification based recommender systems make no distinction between users and accounts; all the data collected during account sessions are attributed to a single user. In reality this is not necessarily true for all accounts; several different users who have distinct, and possibly very different, preferences may access the same account. Such accounts are identified as multi-user accounts. Strangely, no serious study considering the existence of multi-user accounts in recommender systems has been undertaken. This report quantifies the affect multi-user accounts have on the predictive capabilities of recommender system, focusing on two popular collaborative filtering algorithms, the kNN user-based and item-based models. The results indicate that while the item-based model is largely resistant to multi-user account corruption the quality of predictions generated by the user-based model is significantly degraded. / text

Recommender systems have become paramount to customize information access and reduce information overload. They serve multiple uses, ranging from suggesting products and artifacts (to consumers), to bringing people together by the connections induced by (similar) reactions to products and services. This thesis presents a graph-theoretic model that casts recommendation as a process of 'jumping connections' in a graph. In addition to emphasizing the social network aspect, this viewpoint provides a novel evaluation criterion for recommender systems. Algorithms for recommender systems are distinguished not in terms of predicted ratings of services/artifacts, but in terms of the combinations of people and artifacts that they bring together. We present an algorithmic framework drawn from random graph theory and outline an analysis for one particular form of jump called a 'hammock.' Experimental results on two datasets collected over the Internet demonstrate the validity of this approach. / Master of Science

Recommendation is not a new phenomenon arising from the digital era, but an existing social behavior in real life. Recommendation systems facilitate such natural social recommendation behavior and alleviate information overload facing individuals. Among different recommendation techniques proposed in the literature, the collaborative filtering approach is the most successful and widely adopted recommendation technique to date. However, the traditional collaborative filtering recommendation approach ignores proximities between items. That is, all user ratings on items are deemed identically important and given an equal weight in neighborhood formation process. In this study, we proposed a cluster-based collaborative filtering recommendation approach that takes into account the content similarities of items in the collaborative filtering process. Our empirical evaluation results show that the cluster-based collaborative filtering approach improves the prediction accuracy without sacrificing the prediction coverage, using those achieved by the traditional collaborative filtering approach as performance benchmarks. Due to the sparsity problem, when a prediction is made based on few neighbors, the cluster average method could achieve a better prediction accuracy than the proposed approach. Thus, we further proposed an enhanced cluster-based collaborative filtering approach that combines our approach and the cluster average method. The empirical results suggest that the enhanced approach could result in a prediction accuracy comparable to or even better than that accomplished by the cluster average method.

Personal learning environments (PLEs) aim at putting the learner central stage and comprise a technological approach towards learning tools, services, and artifacts gathered from various usage contexts and to be used by learners. Due to the varying technical skills and competences of PLE users, recommendations appear to be useful for empowering learners to set up their environments so that they can connect to learner networks and collaborate on shared artifacts by using the tools available. In this paper we examine different recommender strategies on their applicability in PLE settings. After reviewing different techniques given by literature and experimenting with our prototypic PLE solution we come to the conclusion to start with an item-based strategy and extend it with model-based and iterative techniques for generating recommendations for PLEs. (author's abstract)

Communities of practices are particular and identified knowledge networks involved in a new global, virtual and digital framework. The study of their specific characteristics, the Legitimate Peripheral Participation and the duality Reification/Participation, provides the necessary background to understand and formalize the barriers and the limits in this new context. <br />In order to overcome these ones, the analysis of the tools and the methods for computerized reification (content analysis, information architecture, information visualization) and for enrichment and assessment of content and users (Human-Computer Interactions, Collaborative filtering) enables to develop a methodology to support the application of an automated and interactive reification process in a virtual Communities of Practices.

This dissertation studies the factors that negatively impact the accuracy of the collaborative filtering recommendation systems based on nonnegative matrix factorization (NMF). The keystone in the recommendation system is the rating that expresses the user's opinion about an item. One of the most significant issues in the recommendation systems is the lack of ratings. This issue is called "cold-start" issue, which appears clearly with New-Users who did not rate any item and New-Items, which did not receive any rating. The traditional recommendation systems assume that users are independent and identically distributed and ignore the connections among users whereas the recommendation actually is a social activity. This dissertation aims to enhance NMF-based recommendation systems by utilizing the imputation method and limiting the errors that are introduced in the system. External information such as trust network and item categories are incorporated into NMF-based recommendation systems through the imputation. The proposed approaches impute various subsets of the missing ratings. The subsets are defined based on the total number of the ratings of the user or item before the imputation, such as impute the missing ratings of New-Users, New-Items, or cold-start users or items that suffer from the lack of the ratings. In addition, several factors are analyzed that affect the prediction accuracy when the imputation method is utilized with NMF-based recommendation systems. These factors include the total number of the ratings of the user or item before the imputation, the total number of imputed ratings for each user and item, the average of imputed rating values, and the value of imputed rating values. In addition, several strategies are applied to select the subset of missing ratings for the imputation that lead to increasing the prediction accuracy and limiting the imputation error. Moreover, a comparison is conducted with some popular methods that are in common with the proposed method in utilizing the imputation to handle the lack of ratings, but they differ in the source of the imputed ratings. Experiments on different large-size datasets are conducted to examine the proposed approaches and analyze the effects of the imputation on accuracy. Users and items are divided into three groups based on the total number of the ratings before the imputation is applied and their recommendation accuracy is calculated. The results show that the imputation enhances the recommendation system by capacitating the system to recommend items to New-Users, introduce New-Items to users, and increase the accuracy of the cold-start users and items. However, the analyzed factors play important roles in the recommendation accuracy and limit the error that is introduced from the imputation.

Ever since the rapid growth of the Internet, recommender systems have become essential in helping online users to search and retrieve relevant information they need. Just like the situation that people rely heavily on recommendation in their daily decision making processes, online users may identify desired documents more effectively and efficiently through recommendation of other users who exhibit similar interests, and/or through extracting crucial features of the users¡¦ past preferences. Typical recommendation approaches can be classified into collaborative filtering and content-based filtering. Both approaches, however, have their own drawbacks. The purpose of this research is thus to propose a hybrid approach for text recommendations. We combine collaborative input and document content to facilitate the creation of extended content-based user profiles. These profiles are then rearranged with the technique of latent semantic indexing. Two experiments are conducted to verify our proposed approach. The objective of these experiments is to compare the recommendation results from our proposed approach with those from the other two approaches. The results show that our approach is capable of distinguishing different degrees of document preference, and makes appropriate recommendation to users or does not make recommendation to users for uninterested documents. The application of our proposed approach is justified accordingly.

Στις μέρες μας η χρήση των κινητών συσκευών έχει σημειώσει αλματώδη ανάπτυξη και έχει γίνει αναπόσπαστο κομμάτι της καθημερινότητάς μας. Οι κινητές συσκευές με το πλήθος διαθέσιμων εφαρμογών και δυνατοτήτων που διαθέτουν, καθώς και με τη δυνατότητα πρόσβασης στο Διαδίκτυο, τείνουν να αντικαταστήσουν τους ηλεκτρονικούς υπολογιστές καθώς και μια πληθώρα άλλων συσκευών. Στην παρούσα μεταπτυχιακή διπλωματική εργασία προτείνουμε και υλοποιούμε ένα σύστημα, το οποίο κινείται στα πλαίσια των τεχνολογιών κινητού υπολογισμού και σχεδιάστηκε για να χρησιμοποιείται από τους χρήστες προκειμένου να δέχονται προτάσεις προς επιλογή σχετικά με ταινίες. Το σύστημα αποτελείται από μία κινητή συσκευή η οποία επικοινωνεί με μια οντολογία με χρήση της τεχνολογίας των web services. Όταν ο χρήστης συνδέεται στο λογαριασμό του έχει τη δυνατότητα να πραγματοποιήσει δύο διαδικασίες οι οποίες λειτουργούν με διαφορετικό αλγόριθμο συστάσεων. Στόχος μας είναι να ελέγξουμε κατά πόσο ένα σύστημα συνεργατικών συστάσεων είναι πιο αποδοτικό από ένα σύστημα που λαμβάνει υπόψη το προσωπικό προφίλ ενός χρήστη. Στην παρούσα περίπτωση διαλέξαμε έναν συγκεκριμένο αριθμό ταινιών με παρόμοιο κριτήριο για χρονολογίες από το 2006 έως το 2014. Εφαρμόσαμε τον αλγόριθμο συνεργατικής σύστασης για ταινίες από το 2006 έως το 2010 και τον αλγόριθμο που βασίζεται στο προφίλ μόνο του συνδεδεμένου χρήστη για ταινίες από το 2011 έως το 2014 λαμβάνοντας μια αξιολόγηση για το καθένα. / Nowadays, the use of mobile devices has rapidly developed and has become an integral part of our daily lives. Mobile devices have now a great number of applications and features available, along with the internet accessibility, they tend to replace not only computers but also a variety of other devices. In this master thesis, we propose and implement a system that runs in the context of mobile computing technologies and is designed to be used in order to present to the user all the recommended for him movies. This program consists of a mobile device that communicates with an ontology through a web service. When the user signs in to his account, has the ability to hold two processes each one operating with a different recommendation engine. Our intention is to check whether a collaborative recommendation engine is more efficient than a system which takes into account only the personal profile of a user. In this case study we chose a certain number of films based on a standard for a period of time, from 2006 to 2014. We applied the collaborative recommendation engine to movies from 2006 to 2010 and the algorithm based on the profile of the user signed–into movies released from 2011 to 2014 taking an assessment for each.

Given the proliferation of new-generation internet capable devices in our society, they are now commonly used for a variety of purposes and by a variety of ages, including young children. The vast amount of new media content, available through these devices, cause parents to worry about what their children have access to. In this thesis we investigated how parents and children can work together towards the goal of content control and filtering. One problem to the current content control filtering tools and approaches is that they do not involve children in the filtering process, thus missing an opportunity of educating children about content appropriateness. Therefore, we propose a kid-in-the-loop approach to content control and filtering where parents and children collaboratively configure restrictions and filters, an approach that focuses on education rather than simple rule setting. We conducted an exploratory qualitative study with results highlighting the importance that parents place on avoiding inappropriate content. Building on these findings, we designed an initial kid-in-the-loop prototype which allows parents to work with their children to select appropriate applications, providing parents with the opportunity to educate their children on what they consider to be appropriate or inappropriate. We further validate our proposed approach by conducting a qualitative study with sets of parents and children in the six to eight year-old age group, which revealed an overwhelmingly favorable response to this approach. We conclude this thesis with a comprehensive analysis of our approach, which can be leveraged in designing content control systems targeting both parents and children.