Web and mobile technologies have had such profound impact that we have witnessed significant evolutionary changes in our social, economic and cultural activities. In recent years, online social networking sites such as Twitter, Facebook, Google+, and LinkedIn have gained immense popularity. Such social networks have led to an enormous explosion of network-centric data in a wide variety scenarios, posing unprecedented analytical and computational challenges to MIS researchers. At the same time, the availability of such data offers major research opportunities in various social computing and analytics areas to tackle interesting questions such as: - From a business and marketing perspective, how to mine the novel datasets of online user activities, interpersonal communications and interactions, for developing more successful marketing strategies? - From a system development perspective, how to incorporate massive amounts of available data to assist online users to find relevant, efficient, and timely information? In this dissertation, I explored these research opportunities by studying multiple analytics problems arose from the design and use of social networking services. The first two chapters (Chapter 2 and 3) are intended to study how social network can help to derive a better estimation of customer lifetime value (CLV), in the social gaming context. In Chapter 2, I first conducted an empirical study to demonstrate that friends' activities can serve as significant indicators of a player's CLV. Based on this observation, I proposed a perceptron-based online CLV prediction model considering both individual and friendship information. Preliminary results have shown that the model can be effectively used in online CLV prediction, by evaluating against other commonly-used benchmark methods. In Chapter 3, I further extended the metric of traditional CLV, by incorporating the personal influences on other customers' purchase as an integral part of the lifetime value. The proposed metric was illustrated and tested on seven social games of different genres. The results showed that the new metric can help marketing managers to achieve more successful marketing decisions in user acquisition, user retention, and cross promotion. Chapter 4 is devoted to the design of a recommendation system for micro-blogging. I studied the information diffusion pattern in a micro-blogging site (Twitter.com) and proposed diffusion-based metrics to assess the quality of micro-blogs, and leverage the new metric to implement a novel recommendation framework to help micro-blogging users to efficiently identify quality news feeds. Chapter 5 concludes this dissertation by highlighting major research contributions and future directions.

Increasing number of software vendors are offering or planning to offer their applications as a Software-as-a-Service (SaaS) to leverage the benefits of cloud computing and Internet-based delivery. Therefore, potential clients will face increasing number of providers that satisfy their requirements to choose from. Consequently, there is an increasing demand for automating such a time-consuming and error-prone task. In this work, we develop an architecture for automated service discovery and selection in cloud computing environment. The system is based on an algorithm that recommends service choices to users based on both functional and non-functional characteristics of available services. The system also derives automated ratings from monitoring results of past service invocations to objectively detect badly-behaving providers. We demonstrate the effectiveness of our approach using an early prototype that was developed following object-oriented methodology and implemented using various open-source Java technologies and frameworks. The prototype uses a Chord DHT as its distributed backing store to achieve scalability.

Recommender systems are widely used to achieve a constantly growing variety of services. Alongside with social networks, recommender systems that take into account friendship or trust between users have emerged. In this thesis, we propose an evolution of trust-based recommender systems adapted to decentralized architectures that can be deployed on top of existing social networks. Users profiles are stored locally and are exchanged with a limited, user-defined, list of trusted users. Our approach takes into account friends' similarity and propagates recommendation to direct friends in the social network in order to prevent ratings from being globally known. Moreover, the computational complexity is reduced since calculations are performed on a limited dataset, restricted to the user's neighborhood. On top of this propagation, our approach investigates several aspects. Our system computes and returns to the final user a confidence on the recommendation. It allows the user to tune his/her choice from the recommended products. Confidence takes into account friends' recommendations variance, their number, similarity and freshness of the recommendations. We also propose several heuristics that take into account peer-to-peer constraints, especially regarding network flooding. We show that those heuristics decrease network resources consumption without sacrificing accuracy and coverage. We propose default scoring strategies that are compatible with our constraints. We have implemented and compared our approach with existing ones, using multiple datasets, such as Epinions and Flixster. We show that local information with default scoring strategies are sufficient to cover more users than classical collaborative filtering and trust-based recommender systems. Regarding accuracy, our approach performs better than others, especially for cold start users, even if using less information.

Web services have been developed as an attractive paradigm for publishing, discovering and consuming services. They are loosely-coupled applications that can be run alone or be composed to create new value-added services. They can be consumed as individual services which provide a unique interface to receive inputs and return outputs; or they can be consumed as components to be integrated into business processes. We call the first consumption case individual use and the second case business process use. The requirement of specific tools to assist consumers in the two service consumption cases involves many researches in both academics and industry. On the one hand, many service portals and service crawlers have been developed as specific tools to assist users to search and invoke Web services for individual use. However, current approaches take mainly into account explicit knowledge presented by service descriptions. They make recommendations without considering data that reflect user interest and may require additional information from users. On the other hand, some business process mechanisms to search for similar business process models or to use reference models have been developed. These mechanisms are used to assist process analysts to facilitate business process design. However, they are labor-intense, error-prone, time-consuming, and may make business analyst confused. In our work, we aim at facilitating the service consumption for individual use and business process use using recommendation techniques. We target to recommend users services that are close to their interest and to recommend business analysts services that are relevant to an ongoing designed business process. To recommend services for individual use, we take into account the user's usage data which reflect the user's interest. We apply well-known collaborative filtering techniques which are developed for making recommendations. We propose five algorithms and develop a web-based application that allows users to use services. To recommend services for business process use, we take into account the relations between services in business processes. We target to recommend relevant services to selected positions in a business process. We define the neighborhood context of a service. We make recommendations based on the neighborhood context matching. Besides, we develop a query language to allow business analysts to formally express constraints to filter services. We also propose an approach to extract the service's neighborhood context from business process logs. Finally, we develop three applications to validate our approach. We perform experiments on the data collected by our applications and on two large public datasets. Experimental results show that our approach is feasible, accurate and has good performance in real use-cases

Social endorsements are user-generated endorsements of products or services, such as “likes” and personal collections, in an online social platform. We examine the effect of prior social endorsements on subsequent users’ tendency to endorse or examine a product in a social shopping context, where a social platform connect consumers and enable a collaborative shopping experience. This research consists of two parts. In part I, we identify two ways prior social endorsements can affect subsequent user behavior: as a crowd endorsement, which is an aggregate number of endorsements a product receives for anyone who comes across the product, and as a friend endorsement, which is an endorsement with the endorser’s identity delivered only to the endorser’s friends or followers. Using a panel data of 1656 products on a leading social shopping platform, we quantify the relationship between crowd and friend endorsements and subsequent examination (“click”) and endorsement (“like”) of the products, noting that examination is a private behavior while endorsement is a public behavior. Our results are consistent with the identity signaling theory where identity-conscious consumers converge with the aspiration group (the followers) in their public behavior (e.g. endorsement) and diverge from the avoidance groups (the crowd). We also find differences between public and private behaviors. Moreover, the symbolic nature of social shopping platform trumps the traditional dichotomy of symbolic/functional product attributes. Part II of this study seeks to clarify the underlying mechanism through lab experiments. We hypothesize that consumers’ evaluative attitude, specifically the value-expressive type, moderates the relationship between crowd and friend endorsements and a focal user’s product choice. Our initial results of the second study show support for this idea in the cases when the product choice is not obvious.

Ces dernières années, le contenu disponible sur le Web a augmenté de manière considérable dans ce qu’on appelle communément le Web social. Pour l’utilisateur moyen, il devient de plus en plus difficile de recevoir du contenu de qualité sans se voir rapidement submergé par le flot incessant de publications. Pour les fournisseurs de service, le passage à l’échelle reste problématique. L’objectif de cette thèse est d’aboutir à une meilleure expérience utilisateur à travers la mise en place de systèmes de filtrage et de recommandation. Le filtrage consiste à offrir la possibilité à un utilisateur de ne recevoir qu’un sous ensemble des publications des comptes auxquels il est abonné. Tandis que la recommandation permet la découverte d’information à travers la suggestion de comptes à suivre sur des sujets donnés. Nous avons élaboré MicroFilter un système de filtrage passant à l’échelle capable de gérer des flux issus du Web ainsi que RecLand, un système de recommandation qui tire parti de la topologie du réseau ainsi que du contenu afin de générer des recommandations pertinentes. / In the last years, the amount of available data on the social Web has exploded. For the average user, it became hard to find quality content without being overwhelmed with publications. For service providers, the scalability of such services became a challenging task. The aim of this thesis is to achieve a better user experience by offering the filtering and recommendation features. Filtering consists to provide for a given user, the ability of receiving only a subset of the publications from the direct network. Where recommendation allows content discovery by suggesting relevant content producers on given topics. We developed MicroFilter, a scalable filtering system able to handle Web-like data flows and RecLand, a recommender system that takes advantage of the network topology as well as the content in order to provide relevant recommendations.

Aim The aim of the thesis is to explain the factors that affect the auditor's recommendation concerning audit services to customers who are not subject to mandatory auditing. Background and problem In 2010 mandatory auditing for small companies was abolished. It is common for the auditor to provide recommendations re-garding whether or not a customer should chose to retain the audit. The question is which factors can explain the auditor's recommendation. Method and empirics This thesis uses a deductive approach with inductive elements and a combination of qualitative and quantitative data is used. The qualitative data consists of a pilot study and the quantitative data consists of a questionnaire survey. The analysis of the empirical data was performed using regression analysis. Theory This thesis applies an eclectic approach where the starting point is legitimacy, institutional theory, professional theory and decision making theory to develop a model. Results and conclusions The notion of the recommendation as well as the extent of the recommendations can be explained by factors related to the auditor's agency affiliation and the auditor's personal qualities. / Syfte Studiens syfte är att förklara vilka faktorer som påverkar revisorns rekommendation om revisionstjänster till kunder som inte omfattas av revisionsplikt. Bakgrund och problem År 2010 avskaffades revisionsplikten för mindre bolag. Vid valet om att behålla eller avskaffa revisionen är det vanligt att revisorn rekommenderar kunden om hur bolaget ska välja. Frågan är vilka faktorer som kan förklara revisorns rekommendation. Metod och analys Studien har en deduktiv ansats med induktiva inslag samt an-vänder en kombination av kvalitativ och kvantitativ data. Den kvalitativa datan består utav en pilotstudie och den kvantitativa datan utav en enkätundersökning. Regressionsanalyser genom-fördes vid analys av den empiriska datan. Teori Studien tillämpar ett eklektiskt angreppssätt där utgångspunkten är legitimitet, institutionell teori, professionsteori och beslutsteori som används för att utveckla en modell. Resultat och slutsats Både uppfattning om rekommendation samt omfattningen av rekommendationer kan förklaras av faktorer kopplade till dels revisorns byråtillhörighet och dels revisorns personliga egenskaper.

Studies show that software developers spend about 19% of their development time in web surfing. While collecting necessary information using traditional web search, they face several practical challenges. First, it does not consider context (i.e., surroundings, circumstances) of the programming problems during search unless the developers do so in search query formulation, and forces the developers to frequently switch between their working environment (e.g., IDE) and the web browser. Second, technical details (e.g., stack trace) of an encountered exception often contain a lot of information, and they cannot be directly used as a search query given that the traditional search engines do not support long queries. Third, traditional search generally returns hundreds of search results, and the developers need to manually analyze the result pages one by one in order to extract a working solution. Both manual analysis of a page for content relevant to the encountered exception (and its context) and working an appropriate solution out are non-trivial tasks. Traditional code search engines share the same set of limitations of the web search ones, and they also do not help much in collecting the code examples that can be used for handling the encountered exceptions. In this thesis, we present a context-aware and IDE-based approach that helps one overcome those four challenges above. In our first study, we propose and evaluate a context-aware meta search engine for programming errors and exceptions. The meta search collects results for any encountered exception in the IDE from three popular search engines- Google, Bing and Yahoo and one programming Q & A site- StackOverflow, refines and ranks the results against the detailed context of the encountered exception, and then recommends them within the IDE. From this study, we not only explore the potential of the context-aware and meta search based approach but also realize the significance of appropriate search queries in searching for programming solutions. In the second study, we propose and evaluate an automated query recommendation approach that exploits the technical details of an encountered exception, and recommends a ranked list of search queries. We found the recommended queries quite promising and comparable to the queries suggested by experts. We also note that the support for the developers can be further complemented by post-search content analysis. In the third study, we propose and evaluate an IDE-based context-aware content recommendation approach that identifies and recommends sections of a web page that are relevant to the encountered exception in the IDE. The idea is to reduce the cognitive effort of the developers in searching for content of interest (i.e., relevance) in the page, and we found the approach quite effective through extensive experiments and a limited user study. In our fourth study, we propose and evaluate a context-aware code search engine that collects code examples from a number of code repositories of GitHub, and the examples contain high quality handlers for the exception of interest. We validate the performance of each of our proposed approaches against existing relevant literature and also through several mini user studies. Finally, in order to further validate the applicability of our approaches, we integrate them into an Eclipse plug in prototype--ExcClipse. We then conduct a task-oriented user study with six participants, and report the findings which are significantly promising.

Collaborative filtering based recommender systems use information about a user's preferences to make personalized predictions about content, such as topics, people, or products, that they might find relevant. As the volume of accessible information and active users on the Internet continues to grow, it becomes increasingly difficult to compute recommendations quickly and accurately over a large dataset. In this study, we will introduce an algorithmic framework built on top of Apache Spark for parallel computation of the neighborhood-based collaborative filtering problem, which allows the algorithm to scale linearly with a growing number of users. We also investigate several different variants of this technique including user and item-based recommendation approaches, correlation and vector-based similarity calculations, and selective down-sampling of user interactions. Finally, we provide an experimental comparison of these techniques on the MovieLens dataset consisting of 10 million movie ratings.

With the volume of digital media available today, automatic music recommendation services have proven a useful tool for consumers, allowing them to better discover new and enjoyable music. Typically, this technology is based on collaborative filtering techniques, employing human-generated metadata to base recommendations. Recently, work in content-based recommendation systems have emerged in which the audio signal itself is analyzed for relevant musical information from which models are built that attempt to mimic human similarity judgments. The current state-of-the-art for content-based music recommendation uses a timbre model based on MFCCs calculated on short segments of tracks. These feature vectors are then modeled using GMMs (Gaussian mixture models). GMM modeling of frame-based MFCCs has been shown to perform fairly well on timbre similarity tasks. However, a common problem is that of hubs , in which a relative small number of songs falsely appear similar to many other songs, significantly decreasing the accuracy of similarity recommendations. In this thesis, we explore the origins of hubs in timbre-based modeling and propose several remedies. Specifically, we find that a process of model homogenization, in which certain components of a mixture model are systematically removed, improves performance as measured against several ground-truth similarity metrics. Extending the work of Aucouturier, we introduce several new methods of homogenization. On a subset of the uspop data set, model homogenization improves artist R-precision by a maximum of 3.5% and agreement to user collection co-occurrence data by 7.4%. We also find differences in the effectiveness of the various homogenization methods for hub reduction, with the proposed methods providing the best results. Further, we extend the modeling of frame-based MFCC features by using a kernel density estimation approach to non-parametric modeling. We find that such an approach significantly reduces the number of hubs (by 2.6% of the dataset) while improving agreement to ground-truth by 5% and slightly improving artist R-precision as compared with the standard parametric model. Finally, to test whether these principles hold for all musical data, we introduce an entirely new data set consisting of Indian classical music. We find that our results generalize here as well, suggesting that hubness is a general feature of timbre-based similarity music modeling and that the techniques presented to improve this modeling are effective for diverse types of music.