Current Social Networking Systems (SNS) such as YouTube are creator-driven systems in which creators create content and users search among available content to find what they want. However, queries from users can be time-sensitive, such as some real-time hot topics, which are difficult to obtain at the very moment due to their timeliness and dynamically changing nature. To address this situation, we quest if the system can directly let a user input a query, match the most relevant users (receivers) based on the query and let the receivers decide whether to respond with the very content. In this way, the user can obtain the most relevant data through highly relevant receivers while reducing the reliance on the system's existing data in the recommendation process as an alternative, a new query-driven SNS paradigm. The main objective is to target the most relevant receivers based on a query. In this case, we propose that by allowing users to provide their very moment ideas as queries, the system searches and ranks well-targeted users based on the semantic content of the query and existing user features. However, the user's feature might be incomplete or missing. To alleviate this issue, we propose a novel two-stage query-driven graph-based user recommender system (QDG) that supports query-to-user matching with dynamic update capabilities. In the first stage, we encode the query and item descriptions into attribute features and perform a similarity search to target the Top-N candidate items. In the second stage, we propose a temporal-based graph neural network (t-GNN), which combines the inductive learning-based GNN with the self-attention-based temporal analysis module to predict the most relevant user-item interaction by simultaneously extracting the existing Spatio-temporal features, where spatial feature represents user's relationship with items and temporal feature represents user's behaviour information. We conducted recommendation simulations on six million users and 150,000 merchants on North America YELP data. Experiments show that the QDG system can accurately target strongly relevant users in the North American population based on the query. To the best of our knowledge, we are the first to propose query-driven SNS and demonstrate its effectiveness in a million-scale Yelp dataset.

Users of Social Networking Sites (SNSs) like Facebook, MySpace, LinkedIn, or Twitter face two problems 1) their online social friendships and activities are scattered across SNSs. It is difficult for them to keep track of all their friends and the information about their friends online social activities. 2) they are often overwhelmed by the huge amount of social data (friends updates and other activities). To solve these two problems, this research proposes an approach, named SocConnect. Soc- Connect allows users to create personalized social and semantic contexts for their social data. Users can blend their friends across different social networking sites and group them in different ways. They can also rate friends and/or their activities as favourite, neutral or disliked. SocConnect also can recommend unread friend updates to the user based on user previous ratings on activi- ties and friends, using machine learning techniques. The results from one pilot studies show that users like SocConnects functionalities are needed and liked by the users. An evaluation of the effectiveness of several machine learning algorithms demonstrated that , and machine learning can be usefully applied in predicting the interest level of users in their social network activities, thus helping them deal with the network overload.

Nowadays, successful cooperation and collaboration among developers is crucial to build successful projects in distributed software system development (DSSD). Assigning wrong developers to a specific task not only affects the performance of a component of this task but also affects other components since these projects are composed of dependent components. Another aspect that should be considered when teams are built is the social relationships between the members; disagreements between these members also affect the project team’s performance. These two aspects might cause a project’s failure or delay. Therefore, they are important to consider when teams are created. In this thesis, we developed an Expert Recommender System Framework (ERSF) that assists developers (Active Developers) to find experts who can help them complete or fix the bugs in the code at hand. The ERSF analyzes the developer technical expertise on similar code fragments to the one they need help on assuming that those who have worked on similar fragments might understand and help the Active Developer; also, it analyzes their social relationships with the Active Developer as well as their social activities within the DSSD. Our work is also concerned with improving the system performance and recommendations by tracking the developer communications through our ERSF in order to keep developer profiles up-to-date. Technical expertise and sociality are measured using a combination of technical and social heuristics. The recommender system was tested using scenarios derived from real software development data, and its recommendations compared favourably to recommendations that humans were asked to make in the same scenarios; also, they were compared to the recommendations of the NaiveBayes and other machine learning algorithms. Our experiment results show that ERSF can recommend experts with good to excellent accuracy.

Users of Social Networking Sites (SNSs) like Facebook, MySpace, LinkedIn, or Twitter face two problems 1) their online social friendships and activities are scattered across SNSs. It is difficult for them to keep track of all their friends and the information about their friends online social activities. 2) they are often overwhelmed by the huge amount of social data (friends updates and other activities). To solve these two problems, this research proposes an approach, named SocConnect. Soc- Connect allows users to create personalized social and semantic contexts for their social data. Users can blend their friends across different social networking sites and group them in different ways. They can also rate friends and/or their activities as favourite, neutral or disliked. SocConnect also can recommend unread friend updates to the user based on user previous ratings on activi- ties and friends, using machine learning techniques. The results from one pilot studies show that users like SocConnects functionalities are needed and liked by the users. An evaluation of the effectiveness of several machine learning algorithms demonstrated that , and machine learning can be usefully applied in predicting the interest level of users in their social network activities, thus helping them deal with the network overload.

WiFi-based indoor localization is emerging as a new positioning technology. In this work, we present our efforts to find the best recommender system based on the indoor location tracks collected from the Bow Valley shopping mall for one week. The time a user spends in a shop is considered as an implicit preference and different mapping algorithms are proposed to map the time to a more realistic rating value. A new distribution error metric is proposed to examine the mapping algorithms. Eleven different recommender systems are built and evaluated in terms of accuracy and execution time. The Slope-One recommender system with a logarithmic mapping algorithm is finally selected with a score of 1.292, distribution error of 0.178 and execution time of 0.39 seconds for ten runs.

WiFi-based indoor localization is emerging as a new positioning technology. In this work, we present our efforts to find the best recommender system based on the indoor location tracks collected from the Bow Valley shopping mall for one week. The time a user spends in a shop is considered as an implicit preference and different mapping algorithms are proposed to map the time to a more realistic rating value. A new distribution error metric is proposed to examine the mapping algorithms. Eleven different recommender systems are built and evaluated in terms of accuracy and execution time. The Slope-One recommender system with a logarithmic mapping algorithm is finally selected with a score of 1.292, distribution error of 0.178 and execution time of 0.39 seconds for ten runs.

For learners engaging in senior-level courses, tutors in many cases would like to pick some articles as supplementary reading materials for them each week. Unlike researchers Googling papers from the Internet, tutors, when making recommendations, should consider course syllabus and their assessment of learners along many dimensions. As such, simply Googling articles from the Internet is far from enough. That is, learner models of each individual, including their learning interest, knowledge, goals, etc. should be considered when making paper recommendations, since the recommendation should be carried out so as to ensure that the suitability of a paper for a learner is calculated as the summation of the fitness of the appropriateness of it to help the learner in general. This type of the recommendation is called a Pedagogical Paper Recommender.<p>In this thesis, we propose a set of recommendation methods for a Pedagogical Paper Recommender and study the various important issues surrounding it. Experimental studies confirm that making recommendations to learners in social learning environments is not the same as making recommendation to users in commercial environments such as Amazon.com. In such learning environments, learners are willing to accept items that are not interesting, yet meet their learning goals in some way or another; learners overall impression towards each paper is not solely dependent on the interestingness of the paper, but also other factors, such as the degree to which the paper can help to meet their cognitive goals.<p>It is also observed that most of the recommendation methods are scalable. Although the degree of this scalability is still unclear, we conjecture that those methods are consistent to up to 50 papers in terms of recommendation accuracy. <p>The experiments conducted so far and suggestions made on the adoption of recommendation methods are based on the data we have collected during one semester of a course. Therefore, the generality of results needs to undergo further validation before more certain conclusion can be drawn. These follow up studies should be performed (ideally) in more semesters on the same course or related courses with more newly added papers. Then, some open issues can be further investigated. <p>Despite these weaknesses, this study has been able to reach the research goals set out in the proposed pedagogical paper recommender which, although sounding intuitive, unfortunately has been largely ignored in the research community. <p>Finding a good paper is not trivial: it is not about the simple fact that the user will either accept the recommended items, or not; rather, it is a multiple step process that typically entails the users navigating the paper collections, understanding the recommended items, seeing what others like/dislike, and making decisions. Therefore, a future research goal to proceed from the study here is to design for different kinds of social navigation in order to study their respective impacts on user behavior, and how over time, user behavior feeds back to influence the system performance.

For learners engaging in senior-level courses, tutors in many cases would like to pick some articles as supplementary reading materials for them each week. Unlike researchers Googling papers from the Internet, tutors, when making recommendations, should consider course syllabus and their assessment of learners along many dimensions. As such, simply Googling articles from the Internet is far from enough. That is, learner models of each individual, including their learning interest, knowledge, goals, etc. should be considered when making paper recommendations, since the recommendation should be carried out so as to ensure that the suitability of a paper for a learner is calculated as the summation of the fitness of the appropriateness of it to help the learner in general. This type of the recommendation is called a Pedagogical Paper Recommender.<p>In this thesis, we propose a set of recommendation methods for a Pedagogical Paper Recommender and study the various important issues surrounding it. Experimental studies confirm that making recommendations to learners in social learning environments is not the same as making recommendation to users in commercial environments such as Amazon.com. In such learning environments, learners are willing to accept items that are not interesting, yet meet their learning goals in some way or another; learners overall impression towards each paper is not solely dependent on the interestingness of the paper, but also other factors, such as the degree to which the paper can help to meet their cognitive goals.<p>It is also observed that most of the recommendation methods are scalable. Although the degree of this scalability is still unclear, we conjecture that those methods are consistent to up to 50 papers in terms of recommendation accuracy. <p>The experiments conducted so far and suggestions made on the adoption of recommendation methods are based on the data we have collected during one semester of a course. Therefore, the generality of results needs to undergo further validation before more certain conclusion can be drawn. These follow up studies should be performed (ideally) in more semesters on the same course or related courses with more newly added papers. Then, some open issues can be further investigated. <p>Despite these weaknesses, this study has been able to reach the research goals set out in the proposed pedagogical paper recommender which, although sounding intuitive, unfortunately has been largely ignored in the research community. <p>Finding a good paper is not trivial: it is not about the simple fact that the user will either accept the recommended items, or not; rather, it is a multiple step process that typically entails the users navigating the paper collections, understanding the recommended items, seeing what others like/dislike, and making decisions. Therefore, a future research goal to proceed from the study here is to design for different kinds of social navigation in order to study their respective impacts on user behavior, and how over time, user behavior feeds back to influence the system performance.

Living in the information-overloading age, many people find it difficult to assimilate the information and to identify resources they need. As to a consumer, browsing, searching, and buying a product on online stores is often a time-consuming and frustrating task with the flourishing development of e-commerce. Many shoppers who are interested in buying products on E-commerce websites end up finding nothing they want. Therefore, many E-commerce websites have implemented recommender systems that intend to provide consumers with personalized recommendations for various types of products and services. Some recent research has taken into account social influence in recommender systems in E-commerce. These recommender systems have been observed to achieve better accuracy of prediction, and have also overcome some of the problems of the previous methods. In this study, we propose a trust network-based recommendation framework that utilizes the trust relationship between users to generate recommendation. We employ PageRank algorithm for trust matrix adjustment and recommendation. In addition, we propose several assumptions that can be used to construct trust matrix, and we verify them by experiments. We finally identify two approaches for adjusting trust matrix. Bases on the trust and rating data collected from Epinion.com, we exercise several alternatives and evaluated many combinations of trust matrix adjustment and recommendation methods. Our experiment evaluation results show that using different pagerank for different users groups can generate better recommendation results. Moreover, we proposed a best hybrid method that achieves the best performance.

Twitter, the popular micro-blogging service, has gained a rapid growth in recent years. Newest information is accessible in this social web service through a large volume of real-time tweets. Tweets are short and they are more informative when they are coupled with URLs, which are addresses of interesting web pages related to the tweets. Due to tweet overload in Twitter, an accurate URL recommender system is a bene cial tool for information seekers. In this thesis, we focus on a neighborhoodbased recommender system that recommends URLs to Twitter users. We consider one of the major elements of tweets, hashtags, as the topic representatives of URLs in our approach. We propose methods for incorporating hashtags in measuring the relevancy of URLs. Our experiments show that our neighborhood-based recommender system outperforms a matrix factorization-based system significantly. We also show that the accuracy of URL recommendation in Twitter is time-dependent. A higher recommendation accuracy is obtained when more recent data is provided for recommendation. / Graduate / 0984 / y.nazpar@gmail.com