Effective decision support plays vital roles in people's daily life, as well as for professional practitioners such as health care providers. Without correct information and timely derived knowledge, a decision is often suboptimal and may result in signi cant nancial loss or compromises of the performance. In this dissertation, we study text mining and topic modeling and propose to use text mining methods, in combination with topic models, to discover knowledge from texts popularly available from a wide variety of sources, such as research publications, news, medical diagnose notes, and further employ discovered knowledge to assist social and medical decision support. Examples of such decisions include hospital patient readmission prediction, which is a national initiative for health care cost reduction, academic research topics discovery and trend modeling, and social preference modeling for friend recommendation in social networks etc. To carry out text mining, our research, in Chapter 3, first emphasizes on single document analyzing to investigate textual stylometric features for user pro ling and recognition. Our research confirms that by using properly designed features, it is possible to identify the authors who wrote the article, using a number of sample articles written by the author as the training data. This study serves as the base to assert that text mining is a powerful tool for capturing knowledge in texts for better decision making. In the Chapter 4, we advance our research from single documents to documents with interdependency relationships, and propose to model and predict citation relationship between documents. Given a collection of documents with known linkage relationships, our research will discover e ective features to train prediction models, and predict the likelihood of two documents involving a citation relationships. This study will help accurately model social network linkage relationships, and can be used to assist e ective decision making for friend recommendation in social networking, and reference recommendation in scienti c writing etc. In the Chapter 5, we advance a topic discovery and trend prediction principle to discover meaningful topics from a set of data collection, and further model the evolution trend of the topic. By proposing techniques to discover topics from text, and using temporal correlation between trend for prediction, our techniques can be used to summarize a large collection of documents as meaningful topics, and further forecast the popularity of the topic in a near future. This study can help design systems to discover popular topics in social media, and further assist resource planning and scheduling based on the discovered topics and the their evolution trend. In the Chapter 6, we employ both text mining and topic modeling to the medical domain for effective decision making. The goal is to discover knowledge from medical notes to predict the risk of a patient being re-admitted in a near future. Our research emphasizes on the challenge that re-admitted patients are only a small portion of the patient population, although they bring signficant financial loss. As a result, the datasets are highly imbalanced which often result in poor accuracy for decision making. Our research will propose to use latent topic modeling to carryout localized sampling, and combine models trained from multiple copies of sampled data for accurate prediction. This study can be directly used to assist hospital re-admission assessment for early warning and decision support. The text mining and topic modeling techniques investigated in the dissertation can be applied to many other domains, involving texts and social relationships, towards pattern and knowledge based e ective decision making. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2016. / FAU Electronic Theses and Dissertations Collection

With the increase in bandwidth, more widespread distribution of media, and increased capability of mobile devices, multimedia streaming has not only become feasible, but more economical in terms of space occupied by the media file and the costs involved in attaining it. Although much attention has been paid to peer to peer media streaming over the Internet using HTTP and RTSP, little research has focussed on the use of the Bluetooth protocol for streaming audio and video between mobile devices. This project investigates the feasibility of Bluetooth as a protocol for audio and video streaming between mobile phones using the J2ME platform, through the analysis of Bluetooth protocols, media formats, optimum packet sizes, and the effects of distance on transfer speed. A comparison was made between RFCOMM and L2CAP to determine which protocol could support the fastest transfer speed between two mobile devices. The L2CAP protocol proved to be the most suitable, providing average transfer rates of 136.17 KBps. Using this protocol a second experiment was undertaken to determine the most suitable media format for streaming in terms of: file size, bandwidth usage, quality, and ease of implementation. Out of the eight media formats investigated, the MP3 format provided the smallest file size, smallest bandwidth usage, best quality and highest ease of implementation. Another experiment was conducted to determine the optimum packet size for transfer between devices. A tradeoff was found between packet size and the quality of the sound file, with highest transfer rates being recorded with the MTU size of 668 bytes (136.58 KBps). The class of Bluetooth transmitter typically used in mobile devices (class 2) is considered a weak signal and is adversely affected by distance. As such, the final investigation that was undertaken was aimed at determining the effects of distance on audio streaming and playback. As can be expected, when devices were situated close to each other, the transfer speeds obtained were higher than when devices were far apart. Readings were taken at varying distances (1-15 metres), with erratic transfer speeds observed from 7 metres onwards. This research showed that audio streaming on the J2ME platform is feasible, however using the currently available class of Bluetooth transmitter, video streaming is not feasible. Video files were only playable once the entire media file had been transferred.

This thesis presents an investigation of how to extend the reach of a Bluetooth personal area network by introducing the concept of Bluetooth Hotspots. Currently two Bluetooth devices cannot communicate with each other unless they are within radio range, since Bluetooth is designed as a cable-replacement technology for wireless communications over short ranges. An investigation was done into the feasibility of creating Bluetooth hotspots that allow distant Bluetooth devices to communicate with each other by transporting their communications between these hotspots via an alternative network infrastructure such as an IP network. Two approaches were investigated, masquerading of remote devices by the local hotspot to allow seamless communications and proxying services on remote devices by providing them on a local hotspot using a distributed service discovery database. The latter approach was used to develop applications capable of transporting Bluetooth’s RFCOMM and L2CAP protocols. Quantitative tests were performed to establish the throughput performance and latency of these transport applications. Furthermore, a number of selected Bluetooth services were tested which lead us to conclude that most data-based protocols can be transported by the system.

Dans cette thèse, j'étudie des jeux de données concernant des liens sociaux entre personnes (appels et SMS), leur mobilité ainsi que des informations économiques sur ces personnes, comme leur revenu et leurs dépenses. Les sept travaux couvrent un spectre assez large et apportent des contributions en informatique des réseaux mais aussi en sociologie, économie et géographie. Les questions posées sont très diverses. Comment quantifier la perte d'information causée par une agrégation de flot de liens en série de graphe ? Comment inférer les mouvements des utilisateurs quand on ne connaît que les localisations des utilisateurs aux moments des appels, et que l'on ne détecte donc que les mouvements qui ont eu lieu entre deux appels consécutifs, sans connaître leur nombre ni les instants auxquels ils ont lieu ? Est-il possible de transmettre des SMS dans une région dense en utilisant la densité des téléphones, la mobilité des utilisateurs ainsi que la localité des messages échangés ? Est-il possible de comprendre les inégalités sociales avec une approche Big Data ? Cette dernière question fait l'objet d'une première étude socio-économique approfondie au prisme du Big Data. Il a été possible d'étudier à grande échelle la stratification de la société, l'existence de clubs de riches, la ségrégation spatiale et la structure des dépenses par classe sociale.Au delà de la variété de ces études et de ces nombreuses applications, cette thèse montre que l'analyse de données individuelles riches à l'échelle d'une population permettent de répondre à de nouvelles questions et à d'anciennes hypothèses avec une approche Big Data. Cette thèse tient à mettre l'accent sur la potentialité d'une approche Big Data mais aussi de sa complémentarité avec les approches classiques (modélisation, sociologie avec enquêtes, …). Un effort particulier a été mis dans l'explication des étapes qui amènent aux résultats et dans la prise en compte des biais ce qui est trop souvent négligé. / In this thesis, I have carried out data-driven studies based on rich, large-scale combined data sets including social links between users (calls and SMS), their demographic parameters (age and gender), their mobility and their economic information such as income and spendings. These seven studies bring insights in network science but also in sociology, economy and geography. The questions asked are very diversified. How can one quantify the loss of temporal information caused by the aggregation of link streams into series of graphs? How can one infer mobility of a user from his or her localisations of calls? Is it possible to transmit SMS in a dense region by using the density of phones, the mobility of users and the locality of the messages? How can one quantify and prove empirically the social stratification of the society at a large population scale? I present, for this last question, a first socio-economic study with a data-driven approach. It has been possible to study, at a very large scale, the stratification of the society, the existence of "rich-clubs", the spatial segregation and purchase patterns for each social class. Beyond the variety of studies and their numerous applications, this thesis shows that the analysis of individual rich combined datasets at a large population scale gives the opportunity to answer long-standing hypotheses and to address novel questions. This work not only points out the potentiality of Big Data approach but also its complementarity to classical approaches (modelization, surveys, …). Particular attention was given in order to explain each steps that lead to results and to take into account biases which is too often neglected.