Web usage mining has become popular in various business areas related with Web site development. In Web usage mining, the commonly visited navigational paths are extracted in terms of Web page addresses from the Web server visit logs, and the patterns are used in various applications. The semantic information of the Web page contents is generally not included in Web usage mining. In this thesis, a framework for integrating semantic information with Web usage mining is implemented. The frequent navigational patterns are extracted in the forms of ontology instances instead of Web page addresses and the result is used for making page recommendations to the visitor. Moreover, an evaluation mechanism is implemented to find the success of the recommendation. Test results proved that stronger and more accurate recommendations are obtained by including semantic information in the Web usage mining instead of using on visited Web page addresses.

One-class collaborative filtering is a special type of collaborative filtering methods that aims to deal with datasets that lack counter-examples. In this work, we introduced social networks as a new data source to the one-class collaborative filtering (OCCF) methods and sought ways to benefit from them when dealing with OCCF problems. We divided our research into two parts. In the first part, we proposed different weighting schemes based on social graphs for some well known OCCF algorithms. One of the weighting schemes we proposed outperformed our baselines for some of the datasets we used. In the second part, we focused on the dataset differences in order to find out why our algorithm performed better on some of the datasets. We compared social graphs with the graphs of users and their neighbors generated by the k-NN algorithm. Our research showed that social graphs generated from a specialized domain better improves the recommendation performance than the social graphs generated from a more generic domain.

With the evolution of Web 2.0, most social-networking sites let their members participate in content generation. Users can label items with tags in these websites. A tag can be anything but it is actually a short description of the item. Because tags represent the reason why a user likes an item, but not how much user likes it / they are better identifiers of user profiles than ratings, which are usually numerical values assigned to items by users. Thus, the tag-based contextual representations of music tracks are concentrated in this study. Items are generally represented by vector space models in the content based recommendation systems. In tag-based recommendation systems, users and items are defined in terms of weighted vectors of social tags. When there is a large amount of tags, calculation of the items to be recommended becomes hard, because working with huge vectors is a time-consuming job. The main objective of this thesis is to represent individual tracks (songs) in lower dimensional spaces. An approach is described for creating music recommendations based on user-supplied tags that are augmented with a hierarchical structure extracted for top level genres from Dbpedia. In this structure, each genre is represented by its stylistic origins, typical instruments, derivative forms, sub genres and fusion genres. In addition to very large vector space models, insufficient number of user tags is another problem in the recommendation field. The proposed method is evaluated with different user profiling methods in case of any insufficiency in the number of user tags. User profiles are extended with multi-domain information. By using multi-domain information, the goal of making more successful and realistic predictions is achieved.

The task of the recommendation systems is to recommend items that are relevant to the preferences of users. Two main approaches in recommendation systems are collaborative filtering and content-based filtering. Collaborative filtering systems have some major problems such as sparsity, scalability, new item and new user problems. In this thesis, a hybrid recommendation system that is based on content-boosted collaborative filtering approach is proposed in order to overcome sparsity and new item problems of collaborative filtering. The content-based part of the proposed approach exploits semantic similarities between items based on a priori defined ontology-based metadata in movie domain and derived feature-weights from content-based user models. Using the semantic similarities between items and collaborative-based user models, recommendations are generated. The results of the evaluation phase show that the proposed approach improves the quality of recommendations.

Using page rank and semantic information are frequently used techniques in next page prediction systems. In our work, we extend the use of Page Rank algorithm for next page prediction with several navigational attributes, which are size of the page, duration of the page visit and duration of transition (two page visits sequentially), frequency of page and transition. In our model, we define popularity of transitions and pages by using duration information, use it in a relation with page size, and visit frequency factors. By using the popularity value of pages, we bias conventional Page Rank algorithm and model a next page prediction system that produces page recommendations under given top-n value. Moreover, we extract semantic terms from web URLs in order to tag pages semantically. The extracted terms are mapped into web URLs with different level of details in order to find semantically similar pages for next page recommendations. With this tagging, we model another next page prediction method, which uses Semantic Tagging (ST) similarity and exploits PPR values as a supportive method. Moreover, we model a Hybrid Page Rank (HPR) algorithm that uses both Semantic Tagging based approach and Popularity Based Page Rank values of pages together in order to investigate the effect of PPR and ST with equal weights. In addition, we investigate the effect of local (a synopsis of directed web graph) and global (whole directed web graph) modeling on next page prediction accuracy.

Abstracts This survey is done, based on the medium-large sized insurance company, Kuo-Hwa Life Insurance Company. It is focused on the interactions of its marketing, service quality, insurer¡¦s expected service, perceived service, perceived service quality, customer satisfaction, loyalty, persistency, repurchase and recommendation, in order for the company to evaluate and establish guidelines of the marketing st-rategy. Those who were questionnaired are the insurers over 18-years-old in the metropolitan areas of Tainan. The personnel of Customer Service within the company have distributed 800 questionnaires. Among these, 586 questionnaires were returned with 20 of them void, making it 566 valid. EXCEL, SPSS statistics software were applied to analyze the insured age, marital status, education, annual income and occupation, to better understand tleir perceived service quality, satisfaction, loyalty, persistency and the intention of repurchase of other products from Kuo-Hwa Life Insurance. The research has resulted in: Positive reflection between Perceived Service versus Perceived Service Quality; Positive reflection between Expected Service versus Perceived Service Quality; Positive reflection between Perceived Service versus Satisfaction; Positive reflection between Perceived Service Quality versus Satisfaction; Positive reflection between Expected Service versus Satisfaction; Positive reflection between Satisfaction versus Persistency and Repurchase of Other Insurance Products; Positive reflection between Customer Satisfaction versus recommendation.

Living in an information-overloading age, the original information searching mechanisms are ineffective and impractical. As the e-commerce is more and more popular, using information technology to discover the latent demand of customers becomes an important issue. Hence, a variety of recommendation techniques have been proposed and many of them have been implemented in real systems, mostly in online stores. Among the techniques, the content-based and collaborative filtering approaches are the ones broadly adopted and proved to be successful. Recently, social network-based recommendation approach has been proposed that takes into account the similarities of items with respect to their social closeness. The social network-based approach performs better than content-based approach in some scenarios and it can also avoid recommending articles that have high content similarity to a user¡¦s favorite articles but low quality. Therefore, we propose three hybrid approaches, Switching, Proportional, and Fusion that combine content-based and social network-based approaches in order to achieve a better performance. Our experimental result shows that even though the proposed approaches have pros and cons under different scenarios, in general they achieve better performance than individual approaches. Besides, we generate some synthetic articles that have close content similarities to articles in our collection to evaluate the fidelity of each approach. The experimental results show that approaches incorporating social network information have lower chance to recommend these faked articles.

Background: The School Health Program (SHP) in the Republic of Panama includes preventive healthcare services delivery and referral issuing at elementary schools nationwide. Despite these early prevention efforts, a majority of referrals are apparently not achieving their desired outcome. This idea is supported by the SHP data showing preventable diseases on the rise. Thus, learning the outcome rates of particular referral outcomes may provide a basis for appropriately targeted action. Methods: Three focus groups were conducted with health workers, medical records staff, and teachers, respectively. Following systems thinking and framework analysis, a pilot plan for referral tracking and referral outcome rates obtainment was developed. Finally, the SHP team was surveyed for their perception on the effectiveness and feasibility of the plan, for future implementation. Results: Themes related to referral tracking led directly to the development of a referral tracking pilot plan (RTPP). Survey data analysis revealed that the SHP team perceived the RTPP as an effective way to obtain complete referral tracking and referral outcome rates, and they also found it feasible to implement. Conclusion: Keeping referral records and tracking the SHP referrals is perceived, by those that will be involved in its delivery, as achievable by implementing a RTPP developed from their own recommendations. Once implemented, the resulting obtainment of referral outcome rates may allow them to know if the SHP preventive objective for issuing these referrals is being properly achieved, and to prioritize for targeted action where needed.

Web services increasingly serve as large repositories and conduits of information. However, they do not always allow for the efficient dissemination of this information, particularly in a reactive way. In this report, I describe sALERT, a web-based application that allows for targeted information from various web services to be combined and cross-referenced in order to produce a system that is more convenient and more efficient in reactively disseminating information. This dissemination is performed using mobile notification mechanisms such as text messages, and information targeting is performed using data from social networks and geolocation sources. I present the design, implementation, and plans for future improvement for this service within this report. / text

Σκοπός της εργασίας είναι η μελέτη και εφαρμογή τεχνολογιών σημασιολογικού ιστού για συστήματα συστάσεων, πάνω σε περιεχόμενο που προέρχεται από το διαδίκτυο. Στα πλαίσια της εργασίας σχεδιάστηκε και υλοποιήθηκε διαδικτυακή εφαρμογή που προτείνει άρθρα ειδήσεων λαμβάνοντας υπόψη το προφίλ/ιστορικό του κάθε χρήστη. Λόγω του μεγάλου όγκου πληροφοριών που κατακλύζει το διαδίκτυο συχνά οι χρήστες δυσκολεύονται να ξεχωρίσουν τις πληροφορίες που πραγματικά σχετίζονται με τα ενδιαφέροντα τους. Επιπλέον οι χρήστες έχουν πολύ διαφορετικά ενδιαφέροντα ή προτιμήσεις που μπορούν να ληφθούν υπόψη ώστε να φιλτραριστούν ή να ταξινομηθούν τα αποτελέσματα μιας ερώτησης με σκοπό το αποτέλεσμα να ικανοποιεί τις εξατομικευμένες ανάγκες κάθε χρήστη. Η κατηγορία αυτών των συστημάτων εξατομίκευσης ονομάζεται συστήματα συστάσεων (recommender systems). Τα συστήματα συστάσεων εκμεταλλεύονται τις ιδιαιτερότητες των χρηστών με σκοπό να διευκολύνουν στο να προσδιορίζουν ακριβέστερα τις πληροφορίες ή τις υπηρεσίες για τις οποίες ενδιαφέρονται περισσότερο ή σχετίζονται με τις ανάγκες τους, κάνοντας χρήση ειδικών αλγορίθμων. Οι αλγόριθμοι που χρησιμοποιούνται λαμβάνουν ως είσοδο τα χαρακτηριστικά και τις προτιμήσεις των χρηστών, ή τις σχέσεις μεταξύ των χρηστών ή τα γνωρίσματα των προς σύσταση αντικειμένων και υπολογίζουν το εκτιμώμενο ενδιαφέρον του χρήστη για κάθε αντικείμενο. Στην συνέχεια ταξινομούν ή φιλτράρουν τα αντικείμενα με κριτήριο το εκτιμώμενο ενδιαφέρον. Παρά τη μεγάλη ερευνητική δραστηριότητα στα συστήματα συστάσεων υπάρχουν σημαντικά προβλήματα που δεν έχουν λυθεί ακόμα πλήρως και απαιτείται περαιτέρω έρευνα. Για παράδειγμα οι τυπικές προσεγγίσεις εξαρτώνται από το πεδίο ορισμού(domain). Τα μοντέλα τους δημιουργούνται από τις πληροφορίες που συλλέγονται μέσα σε ένα συγκεκριμένο πεδίο(domain), και δεν μπορούν να επεκταθούν ή να ενσωματωθούν σε άλλα συστήματα. Επιπλέον η ανάγκη για περαιτέρω ευελιξία με τη μορφή συστάσεων που εξάγονται από επερωτήσεις ή προτάσεων που προσανατολίζονται σε ομάδες χρηστών, καθώς και η εξέταση πλαισιακών χαρακτηριστικών στη διάρκεια των διαδικασιών δημιουργίας συστάσεων είναι και αυτές απαιτήσεις που δεν πληρούνται στα περισσότερα συστήματα. Στην εργασία αυτή παρουσιάζουμε ένα σύστημα συστάσεων που χρησιμοποιεί τεχνολογίες σημασιολογικού ιστού για να περιγράψει και να συνδέσει τις ειδήσεις με τις προτιμήσεις του χρήστη ώστε να δημιουργήσει βελτιωμένες συστάσεις. Οι περιγραφές των ειδήσεων και τα προφίλ των χρηστών δημιουργούνται με την βοήθεια εννοιών που ορίζονται σε ένα σύνολο οντολογιών πεδίου. Ανάλογα με τις ομοιότητες μεταξύ των περιγραφών των ειδήσεων και των προφίλ των χρηστών καθώς και τις σημασιολογικές σχέσεις μεταξύ των εννοιών, το σύστημα υποστηρίζει μοντέλα συστάσεων βάσει περιεχομένου που έχουν σαν επίκεντρο το μεμονωμένο χρήστη, και επιτρέπει την εξαγωγή συμπερασμάτων βασισμένα σε κανόνες για την υποστήριξη εξατομικευμένων συστάσεων. Συγκεκριμένα γίνεται αξιολόγηση του μοντέλου που εξατομικεύει τη σειρά με την οποία τα άρθρα ειδήσεων παρουσιάζονται στο χρήστη λαμβάνοντας υπόψη το προφίλ/ιστορικό των βραχυπρόθεσμων και των μακροπρόθεσμων ενδιαφερόντων. / The scope of this Msc Thesis is the study and applies Semantic Web Technologies, for Recommendation Systems, over content for the internet. For the purpose of work, we designed and implemented web application that proposes news articles considering the profile/ history of each user. Because of the information overload which invading the internet, often the users are complicated to distinguish the information that really is related to their interests. The category of these personalization systems called recommendation systems. More over the users have very different interests or preferences that can taken into account in order to classify or filtering the results of question with scope the result to satisfies the personalized needs of each user. The category of these personalization systems called recommendation systems. Recommendation systems exploit the particularities of users with scope facilitate to identify precisely the information or the services for which they are more interested or related to their needs, using special algorithms. The algorithms used take as input the attributes and the user’s preferences, or the relations between users or the attributes of the items to be recommender and calculate the estimated interest of user for each item. Then classify or filtering the items with criterion the estimated interest. Despite the great research activity in recommendation systems common problem have not fully solved yet, and further investigation is needed. For example, typical approach dependent from domain. The model are created from the information where collected in specific domain, and cannot be extended or integrated in other systems. More over the need for further flexibility in the recommendation derived from question or oriented recommendation to group users, and the consideration of contextual features during the recommendation process are also unfulfilled requirements in most systems. This thesis presents news recommendations systems which used semantic web technologies to describe and relate news items, and the user preferences in order to produce enhanced recommendations. The items descriptions and the user profiles are created with concepts in the domain ontology. According to the similarity between the description items and the user profiles, and the semantic relation between concepts, the system supported content –based model that centered on a single user, and allows the Inference rule-based for the supported personalized recommendation. Specifically an evaluation of the model that personalized the order in which news articles are presented to the user, considering the profile/ history of sort – terms and long – terms interests.