In this thesis a number of models for recommender systems are explored, all using collaborative filtering to produce their recommendations. Extra focus is put on two models: Matrix Factorization, which is a linear model and Multi-Layer Perceptron, which is a non-linear model. With an additional purpose of training the models without collecting any sensitive data from the users, both models were implemented with a learning technique that does not require the server's knowledge of the users' data, called federated learning. The federated version of Matrix Factorization is already well-researched, and has proven not to protect the users' data at all; the data is derivable from the information that the users communicate to the server that is necessary for the learning of the model. However, on the federated Multi-Layer Perceptron model, no research could be found. In this thesis, such a model is therefore designed and presented. Arguments are put forth in support of the privacy preservability of the model, along with a proof of the user data not being analytically derivable for the central server.    In addition, new ways to further put the protection of the users' data on the test are discussed. All models are evaluated on two different data sets. The first data set contains data on ratings of movies and is called MovieLens 1M. The second is a data set that consists of anonymized fund transactions, provided by the Swedish bank SEB for this thesis. Test results suggest that the federated versions of the models can achieve similar recommendation performance as their non-federated counterparts.

Cette thèse porte sur l'étude des Systèmes de Recommandation dans un cadre séquentiel, où les retours des utilisateurs sur des articles arrivent dans le système l'un après l'autre. Après chaque retour utilisateur, le système doit le prendre en compte afin d'améliorer les recommandations futures. De nombreuses techniques de recommandation ou méthodologies d'évaluation ont été proposées par le passé pour les problèmes de recommandation. Malgré cela, l'évaluation séquentielle, qui est pourtant plus réaliste et se rapproche davantage du cadre d'évaluation d'un vrai système de recommandation, a été laissée de côté. Le contexte séquentiel nécessite de prendre en considération différents aspects non visibles dans un contexte fixe. Le premier de ces aspects est le dilemme dit d'exploration vs. exploitation: le modèle effectuant les recommandations doit trouver le bon compromis entre recueillir de l'information sur les goûts des utilisateurs à travers des étapes d'exploration, et exploiter la connaissance qu'il a à l'heure actuelle pour maximiser le feedback reçu. L'importance de ce premier point est mise en avant à travers une première évaluation, et nous proposons une approche à la fois simple et efficace, basée sur la Factorisation de Matrice et un algorithme de Bandit Manchot, pour produire des recommandations appropriées. Le second aspect pouvant apparaître dans le cadre séquentiel surgit dans le cas où une liste ordonnée d'articles est recommandée au lieu d'un seul article. Dans cette situation, le feedback donné par l'utilisateur est multiple: la partie explicite concerne la note donnée par l'utilisateur concernant l'article choisi, tandis que la partie implicite concerne les articles cliqués (ou non cliqués) parmi les articles de la liste. En intégrant les deux parties du feedback dans un modèle d'apprentissage, nous proposons une approche basée sur la Factorisation de Matrice, qui peut recommander de meilleures listes ordonnées d'articles, et nous évaluons cette approche dans un contexte séquentiel particulier pour montrer son efficacité. / This thesis is dedicated to the study of Recommendation Systems under a sequential setting, where the feedback given by users on items arrive one after another in the system. After each feedback, the system has to integrate it and try to improve future recommendations. Many techniques or evaluation methods have already been proposed to study the recommendation problem. Despite that, such sequential setting, which is more realistic and represent a closer framework to a real Recommendation System evaluation, has surprisingly been left aside. Under a sequential context, recommendation techniques need to take into consideration several aspects which are not visible for a fixed setting. The first one is the exploration-exploitation dilemma: the model making recommendations needs to find a good balance between gathering information about users' tastes or items through exploratory recommendation steps, and exploiting its current knowledge of the users and items to try to maximize the feedback received. We highlight the importance of this point through the first evaluation study and propose a simple yet efficient approach to make effective recommendation, based on Matrix Factorization and Multi-Armed Bandit algorithms. The second aspect emphasized by the sequential context appears when a list of items is recommended to the user instead of a single item. In such a case, the feedback given by the user includes two parts: the explicit feedback as the rating, but also the implicit feedback given by clicking (or not clicking) on other items of the list. By integrating both feedback into a Matrix Factorization model, we propose an approach which can suggest better ranked list of items, and we evaluate it in a particular setting.

推薦系統為網站提升使用者滿意度、減少使用者所花費的時間並且替網站提供方提升銷售，是現在網站中不可或缺的要素，而推薦系統的研究集中在娛樂項目，學術研究論文推薦系統的研究有限。若能給予有價值的相關文獻，提供協助，無疑是加速進步的速度。 在過去的研究中，為了達到個人化目的所使用的方法，都有不可避免或未解決的缺點，2002年美國研究圖書館協會提出布達佩斯開放獲取計劃(Budapest Open Access Initiative)，不要求使用者註冊帳號與支付款項就能取得研究論文全文，這樣的做法使期刊走向開放的風氣開始盛行，時至今日，開放獲取對學術期刊網站帶來重大的影響。在這樣的時空背景之下，本研究提出一個適用於學術論文之推薦機制，以FP-Growth演算法與協同過濾做為推薦方法的基礎，消弭過去研究之缺點，並具個人化推薦的優點，經實驗驗證後，證實本研究所提出的推薦架構具有良好的成效。 / Recommendation system is used in many field like movie, music, electric commerce and library. It’s not only save customers’ time but also raise organizations’ efficient. Recommended system is an essential element in a website. Some methods have been developed for recommended system, but they are primarily focused on content or collaboration-based mechanisms. For academic research, it is very important that relevant literature can be provided to researchers when they conduct literature review. Previous research indicates that there are inevitable or unsolved shortcomings in existing methods such as cold starts. Association of Research Libraries purpose “Budapest Open Access Initiative” that is advocate open access concept. Open access means that users can get full paper without register and pay fee. It’s a major impact to academic journal website. In this space-time background, we propose a hybrid recommendation mechanism that takes into consideration the nature of recommendation academic papers to mitigate the shortcomings of existing methods.

Submitted by Johnny Rodrigues (johnnyrodrigues@ufcg.edu.br) on 2018-04-11T14:50:08Z No. of bitstreams: 1 CAIO SANTOS BEZERRA NÓBREGA - DISSERTAÇÃO PPGCC 2014..pdf: 983246 bytes, checksum: 5eca7651706ce317dc514ec2f1aa10c3 (MD5) / Made available in DSpace on 2018-04-11T14:50:08Z (GMT). No. of bitstreams: 1 CAIO SANTOS BEZERRA NÓBREGA - DISSERTAÇÃO PPGCC 2014..pdf: 983246 bytes, checksum: 5eca7651706ce317dc514ec2f1aa10c3 (MD5) Previous issue date: 2014-07-30 / Capes / Sugerir os produtos mais apropriados aos diversos tipos de consumidores não é uma tarefa trivial, apesar de ser um fator chave para aumentar satisfação e lealdade destes. Devido a esse fato, sistemas de recomendação têm se tornado uma ferramenta importante para diversas aplicações, tais como, comércio eletrônico, sites personalizados e redes sociais. Recentemente, a fatoração de matrizes se tornou a técnica mais bem sucedida de implementação de sistemas de recomendação. Os parâmetros do modelo de fatoração de matrizes são tipicamente aprendidos por meio de métodos numéricos, tal como o gradiente descendente. O desempenho do gradiente descendente está diretamente relacionada à configuração da taxa de aprendizagem, a qual é tipicamente configurada para valores pequenos, com o objetivo de não perder um mínimo local. Consequentemente, o algoritmo pode levar várias iterações para convergir. Idealmente,é desejada uma taxa de aprendizagem que conduza a um mínimo local nas primeiras iterações, mas isto é muito difícil de ser realizado dada a alta complexidade do espaço de valores a serem pesquisados. Começando com um estudo exploratório em várias bases de dados de sistemas de recomendação, observamos que, para a maioria das bases, há um padrão linear entre a taxa de aprendizagem e o número de iterações necessárias para atingir a convergência. A partir disso, propomos utilizar modelos de regressão lineares simples para predizer, para uma base de dados desconhecida, um bom valor para a taxa de aprendizagem inicial. A ideia é estimar uma taxa de aprendizagem que conduza o gradiente descendenteaummínimolocalnasprimeirasiterações. Avaliamosnossatécnicaem8bases desistemasderecomendaçãoreaisecomparamoscomoalgoritmopadrão,oqualutilizaum valorfixoparaataxadeaprendizagem,ecomtécnicasqueadaptamataxadeaprendizagem extraídas da literatura. Nós mostramos que conseguimos reduzir o número de iterações até em 40% quando comparados à abordagem padrão. / Suggesting the most suitable products to different types of consumers is not a trivial task, despite being a key factor for increasing their satisfaction and loyalty. Due to this fact, recommender systems have be come an important tool for many applications, such as e-commerce, personalized websites and social networks. Recently, Matrix Factorization has become the most successful technique to implement recommendation systems. The parameters of this model are typically learned by means of numerical methods, like the gradient descent. The performance of the gradient descent is directly related to the configuration of the learning rate, which is typically set to small values, in order to do not miss a local minimum. As a consequence, the algorithm may take several iterations to converge. Ideally, one wants to find a learning rate that will lead to a local minimum in the early iterations, but this is very difficult to achieve given the high complexity of search space. Starting with an exploratory study on several recommendation systems datasets, we observed that there is an over all linear relationship between the learnin grate and the number of iterations needed until convergence. From this, we propose to use simple linear regression models to predict, for a unknown dataset, a good value for an initial learning rate. The idea is to estimate a learning rate that drives the gradient descent as close as possible to a local minimum in the first iteration. We evaluate our technique on 8 real-world recommender datasets and compared it with the standard Matrix Factorization learning algorithm, which uses a fixed value for the learning rate over all iterations, and techniques fromt he literature that adapt the learning rate. We show that we can reduce the number of iterations until at 40% compared to the standard approach.

La diversité des contenus recommandation et la variation des contextes des utilisateurs rendent la prédiction en temps réel des préférences des utilisateurs de plus en plus difficile mettre en place. Toutefois, la plupart des approches existantes n'utilisent que le temps et l'emplacement actuels séparément et ignorent d'autres informations contextuelles sur lesquelles dépendent incontestablement les préférences des utilisateurs (par exemple, la météo, l'occasion). En outre, ils ne parviennent pas considérer conjointement ces informations contextuelles avec les interactions sociales entre les utilisateurs. D'autre part, la résolution de problèmes classiques de recommandation (par exemple, aucun programme de télévision vu par un nouvel utilisateur connu sous le nom du problème de démarrage froid et pas assez d'items co-évalués par d'autres utilisateurs ayant des préférences similaires, connu sous le nom du problème de manque de donnes) est d'importance significative puisque sont attaqués par plusieurs travaux. Dans notre travail de thèse, nous proposons un modèle probabiliste qui permet exploiter conjointement les informations contextuelles actuelles et l'influence sociale afin d'améliorer la recommandation des items. En particulier, le modèle probabiliste vise prédire la pertinence de contenu pour un utilisateur en fonction de son contexte actuel et de son influence sociale. Nous avons considérer plusieurs éléments du contexte actuel des utilisateurs tels que l'occasion, le jour de la semaine, la localisation et la météo. Nous avons utilisé la technique de lissage Laplace afin d'éviter les fortes probabilités. D'autre part, nous supposons que l'information provenant des relations sociales a une influence potentielle sur les préférences des utilisateurs. Ainsi, nous supposons que l'influence sociale dépend non seulement des évaluations des amis mais aussi de la similarité sociale entre les utilisateurs. Les similarités sociales utilisateur-ami peuvent être établies en fonction des interactions sociales entre les utilisateurs et leurs amis (par exemple les recommandations, les tags, les commentaires). Nous proposons alors de prendre en compte l'influence sociale en fonction de la mesure de similarité utilisateur-ami afin d'estimer les préférences des utilisateurs. Nous avons mené une série d'expérimentations en utilisant un ensemble de donnes réelles issues de la plateforme de TV sociale Pinhole. Cet ensemble de donnes inclut les historiques d'accès des utilisateurs-vidéos et les réseaux sociaux des téléspectateurs. En outre, nous collectons des informations contextuelles pour chaque historique d'accès utilisateur-vidéo saisi par le système de formulaire plat. Le système de la plateforme capture et enregistre les dernières informations contextuelles auxquelles le spectateur est confronté en regardant une telle vidéo.Dans notre évaluation, nous adoptons le filtrage collaboratif axé sur le temps, le profil dépendant du temps et la factorisation de la matrice axe sur le réseau social comme tant des modèles de référence. L'évaluation a port sur deux tâches de recommandation. La première consiste sélectionner une liste trie de vidéos. La seconde est la tâche de prédiction de la cote vidéo. Nous avons évalué l'impact de chaque élément du contexte de visualisation dans la performance de prédiction. Nous testons ainsi la capacité de notre modèle résoudre le problème de manque de données et le problème de recommandation de démarrage froid du téléspectateur. Les résultats expérimentaux démontrent que notre modèle surpasse les approches de l'état de l'art fondes sur le facteur temps et sur les réseaux sociaux. Dans les tests des problèmes de manque de donnes et de démarrage froid, notre modèle renvoie des prédictions cohérentes différentes valeurs de manque de données. / Due to the diversity of alternative contents to choose and the change of users' preferences, real-time prediction of users' preferences in certain users' circumstances becomes increasingly hard for recommender systems. However, most existing context-aware approaches use only current time and location separately, and ignore other contextual information on which users' preferences may undoubtedly depend (e.g. weather, occasion). Furthermore, they fail to jointly consider these contextual information with social interactions between users. On the other hand, solving classic recommender problems (e.g. no seen items by a new user known as cold start problem, and no enough co-rated items with other users with similar preference as sparsity problem) is of significance importance since it is drawn by several works. In our thesis work, we propose a context-based approach that leverages jointly current contextual information and social influence in order to improve items recommendation. In particular, we propose a probabilistic model that aims to predict the relevance of items in respect with the user's current context. We considered several current context elements such as time, location, occasion, week day, location and weather. In order to avoid strong probabilities which leads to sparsity problem, we used Laplace smoothing technique. On the other hand, we argue that information from social relationships has potential influence on users' preferences. Thus, we assume that social influence depends not only on friends' ratings but also on social similarity between users. We proposed a social-based model that estimates the relevance of an item in respect with the social influence around the user on the relevance of this item. The user-friend social similarity information may be established based on social interactions between users and their friends (e.g. recommendations, tags, comments). Therefore, we argue that social similarity could be integrated using a similarity measure. Social influence is then jointly integrated based on user-friend similarity measure in order to estimate users' preferences. We conducted a comprehensive effectiveness evaluation on real dataset crawled from Pinhole social TV platform. This dataset includes viewer-video accessing history and viewers' friendship networks. In addition, we collected contextual information for each viewer-video accessing history captured by the plat form system. The platform system captures and records the last contextual information to which the viewer is faced while watching such a video. In our evaluation, we adopt Time-aware Collaborative Filtering, Time-Dependent Profile and Social Network-aware Matrix Factorization as baseline models. The evaluation focused on two recommendation tasks. The first one is the video list recommendation task and the second one is video rating prediction task. We evaluated the impact of each viewing context element in prediction performance. We tested the ability of our model to solve data sparsity and viewer cold start recommendation problems. The experimental results highlighted the effectiveness of our model compared to the considered baselines. Experimental results demonstrate that our approach outperforms time-aware and social network-based approaches. In the sparsity and cold start tests, our approach returns consistently accurate predictions at different values of data sparsity.

Mängden data som transporteras över Internet idag är stor. Vilket innebär att det finns ett överflöd av information och ett behov för att urskilja relevant innehåll mot irrelevant. För att uppnå detta används rekommendationssystem som i sin tur använder olika filtreringstekniker. Det finns tre huvudtyper av tekniker, innehållsbaserad kollaborativ filtrering och hybrid tekniker. Syftet med studien är att jämföra olika filtreringstekniker och undersöka hur teknikerna förhåller sig till träffsäkerhet mot testset, beräkningsbelasning och användarnöjdhet. För att genomföra detta på ett nyanserat sätt har en strukturerad litteraturstudie genomförts där sju olika steg för inhämtning och analys av dokument gjorts. En kvalitativ metanalys genomfördes på de 28 utvalda tidskriftsartiklarna. IFT rekommendationssystem var den rekommendationsteknik som anses minst effektiv utifrån den definition som studien utgått från. KFT och hybrid rekommendationssystem är de som är mest effektiva enligt denna studie. Hybrid system kan ta vara på fördelar från alla andra tekniker och kan synergiskt också motverka vissa av teknikernas förknippade brister, mot kostnaden av att implementations-komplexiteten ökar.

Conselho Nacional de Desenvolvimento Científico e Tecnológico / Hoje em dia, a quantidade de usuários que utilizam sites de comércio eletrônico para realizar compras está aumentando muito, principalmente devido à facilidade e rapidez. Muitos sites de comércio eletrônico, diferentemente das lojas físicas, disponibilizam aos seus usuários uma grande variedade de produtos e serviços, e os usuários podem ter muita dificuldade em encontrar produtos de sua preferência. Normalmente, a preferência por um produto pode ser influenciada pela aparência visual da imagem do produto. Neste contexto, os Sistemas de Recomendação de produtos que estão associados a Imagens (SRI) tornaram-se indispensáveis para ajudar os usuários a encontrar produtos que podem ser, eventualmente, agradáveis ou úteis para eles. Geralmente, os SRI usam o comportamento passado dos usuários (cliques, compras, críticas, avaliações, etc.) e/ou atributos de produtos para definirem as preferências dos usuários. Um dos principais desafios enfrentados em SRI é a necessidade de o usuário fornecer algumas informações sobre suas preferências sobre os produtos, a fim de obter novas recomendações do sistema. Infelizmente, os usuários nem sempre estão dispostos a fornecer tais informações de forma explícita. Assim, a fim de lidar com esse desafio, os métodos para obtenção de informações de forma implícita do usuário são desejáveis. Neste trabalho, propõe-se investigar em que medida informações sobre atenção visual do usuário podem ajudar a melhorar a predição de avaliação e consequentemente produzir SRI mais precisos. É também objetivo deste trabalho o desenvolvimento de dois novos métodos, um método baseado em Filtragem Colaborativa (FC) que combina avaliações e dados de atenção visual para representar o comportamento passado dos usuários, e outro método baseado no conteúdo dos itens, que combina atributos textuais, características visuais e dados de atenção visual para compor o perfil dos itens. Os métodos propostos foram avaliados em uma base de imagens de pinturas e uma base de imagens de roupas. Os resultados experimentais mostram que os métodos propostos neste trabalho possuem ganhos significativos em predição de avaliação e precisão na recomendação quando comparados ao estado-da-arte. Vale ressaltar que as técnicas propostas são flexíveis, podendo ser utilizadas em outros cenários que exploram a atenção visual dos itens recomendados. / Nowadays, the amount of users using e-commerce sites for shopping is greatly increasing, mainly due to the easiness and rapidity of this way of consumption. Many e-commerce sites, differently from physical stores, can offer their users a wide range of products and services, and the users can find it very difficult to find products of your preference. Typically, your preference for a product can be influenced by the visual appearance of the product image. In this context, Image Recommendation Systems (IRS) have become indispensable to help users to find products that may possibly pleasant or be useful to them. Generally, IRS use past behavior of users (clicks, purchases, reviews, ratings, etc.) and/or attributes of the products to define the preferences of users. One of the main challenges faced by IRS is the need of the user to provide some information about his / her preferences on products in order to get further recommendations from the system. Unfortunately, users are not always willing to provide such information explicitly. So, in order to cope with this challenge, methods for obtaining user’s implicit feedback are desirable. In this work, the author propose an investigation to discover to which extent information concerning user visual attention can help improve the rating prediction hence produce more accurate IRS. This work proposes to develop two new methods, a method based on Collaborative Filtering (CF) which combines ratings and data visual attention to represent the past behavior of users, and another method based on the content of the items, which combines textual attributes, visual features and visual attention data to compose the profile of the items. The proposed methods were evaluated in a painting dataset and a clothing dataset. The experimental results show significant improvements in rating prediction and precision in recommendation when compared to the state-of-the-art. It is worth mentioning that the proposed techniques are flexible and can be applied in other scenarios that exploits the visual attention of the recommended items. / Tese (Doutorado)

After decades of black-boxing the existence of algorithms in technologies of daily need, users lack confidence in handling them. This thesis study investigates the use situation of intelligent music recommendation systems and explores how understandability as a principle drawn from sociology, design, and computing can enhance the algorithmic experience. In a Research-Through-Design approach, the project conducted focus user sessions and an expert interview to explore first-hand insights. The analysis showed that users had limited mental models so far but brought curiosity to learn. Explorative prototyping revealed that explanations could improve the algorithmic experience in music recommendation systems. Users could comprehend information the best when it was easy to access and digest, directly related to user behavior, and gave control to correct the algorithm. Concluding, trusting users with more transparent handling of algorithmic workings might make authentic recommendations from intelligent systems applicable in the long run.

Les systèmes de recommandation conversationnels (CRSs) sont des systèmes qui fournissent des recommandations personnalisées par le biais d’une session de dialogue en langage naturel avec les utilisateurs. Contrairement aux systèmes de recommandation traditionnels qui ne prennent comme vérité de base que les préférences anciennes des utilisateurs, les CRS impliquent aussi les préférences actuelles des utilisateurs durant la conversation. Des recherches récentes montrent que la compréhension de la signification contextuelle des préférences des utilisateurs et des dialogues peut améliorer de manière significative les performances du système de recommandation. Des chercheurs ont également montré un lien fort entre les traits de personnalité des utilisateurs et les systèmes de recommandation. La personnalité et les préférences sont des variables essentielles en sciences sociales. Elles décrivent les différences entre les personnes, que ce soit au niveau individuel ou collectif. Les approches récentes de recommandation basées sur la personnalité sont des systèmes non conversationnels. Par conséquent, il est extrêmement important de détecter et d’utiliser les traits de personnalité des individus dans les systèmes conversationnels afin d’assurer une performance de recommandation et de dialogue plus personnalisée. Pour ce faire, ce travail propose un système de recommandation conversationnel sensible à la personnalité qui est basé sur des modules qui assurent une session de dialogue et recommandation personnalisée en utilisant les traits de personnalité des utilisateurs. Nous proposons également une nouvelle approche de détection de la personnalité, qui est un modèle de langage spécifique au contexte pour détecter les traits des individus en utilisant leurs données publiées sur les réseaux sociaux. Les résultats montrent que notre système proposé a surpassé les approches existantes dans différentes mesures. / A Conversational Recommendation System (CRS) is a system that provides personalized recommendations through a session of natural language dialogue turns with users. Unlike traditional one-shot recommendation systems, which only assume the user’s previous preferences as the ground truth, CRS uses both previous and current user preferences. Recent research shows that understanding the contextual meaning of user preferences and dialogue turns can significantly improve recommendation performance. It also shows a strong link between users’ personality traits and recommendation systems. Personality and preferences are essential variables in computational sociology and social science. They describe the differences between people, both at the individual and collective level. Recent personality-based recommendation approaches are traditional one-shot systems, or “non conversational systems”. Therefore, there is a significant need to detect and employ individuals’ personality traits within the CRS paradigm to ensure a better and more personalized dialogue recommendation performance. Driven by the aforementioned facts, this study proposes a modularized, personality- aware CRS that ensures a personalized dialogue recommendation session using the users’ personality traits. We also propose a novel personality detection approach, which is a context-specific language model for detecting individuals’ personality traits using their social media data. The goal is to create a personality-aware and topic-guided CRS model that performs better than the standard CRS models. Experimental results show that our personality-aware conversation recommendation system has outperformed state-of-the-art approaches in different considered metrics on the topic-guided conversation recommendation dataset.

[pt] Os sites de redes sociais ganharam importância nos últimos anos. Neles, seus usuários podem se conectar com outros usuários para interagir entre si. Porém, geralmente o número de usuários cadastrados é muito grande e, por isso, encontrar outros usuários afins para se conectar não é fácil. Sistemas de recomendação são ferramentas de software que provem sugestões de diversos tipos de itens para usuários e podem ser aplicados para recomendar pessoas (outros usuários) em redes sociais. Sistemas que recomendam pessoas utilizam técnicas específicas e, devido às implicações sociais envolvidas nas relações pessoais, devem levar alguns fatores em consideração. A ausência de dados confiáveis torna a tarefa de gerar recomendações úteis mais difícil. Este trabalho discute o assunto e apresenta um sistema de recomendação de pessoas para sites de redes sociais com base em conexões entre usuários. Para testar o sistema apresentado, realizamos um experimento com o Peladeiro, um site real de uma rede social que conta com mais de 500 mil usuários, onde poucos dados estão disponíveis para serem utilizados. / [en] Social networking websites have gained importance in recent years. In them, users can connect with other users to interact with. However, generally, the number of registered users is very large. Therefore, find other users with common interests is not easy. Recommender systems are software tools which generate suggestions for various types of items to users and can be applied to recommend people (other users) on social networks. Systems that recommend people use specific techniques and, due to the social implications involved in personal relationships, must take several factors into consideration. The lack of data available makes the task of generate good recommendations more difficult. This paper discusses the theme and presents a person recommendation system for social networking websites based in user connections. To test the system presented, we conducted an experiment with Peladeiro, a real website of a social network that has over 500 000 users, where few reliable data are available.