Sharif Digital Repository

Recommender systems are in the center of network science and becoming increasingly important in individual businesses for providing efficient personalized services and products to users. The focus of previous research in the field of recommendation systems was on improving the precision of the system through designing more accurate recommendation lists. Recently, the community has been paying attention to diversity and novelty of recommendation list as key characteristics of modern recommender systems. In many cases, novelty and precision do not go in the same direction and the accuracy-novelty dilemma is one of the challenging problems in recommender systems, which needs efforts in making a... 

Recommending routes to city taxis in order to improve their performance in finding passengers can reduce traffic, pollution and waiting time for passengers, as well as this, increasing the efficiency of taxis can increase their income. In this study, by analyzing trajectory data obtained from 536 taxis in San Francisco over a period of one month, we generated networks of cells, each of which has time-dependent characteristics such as the probability of finding a passenger, the capacity of each cell, or the amount of demand exists is in the cell and the average speed of passing through that cell. also, the connections between these cells in order to make more use of the experiences of taxi... 

The World Wide Web has been experiencing a massive growth regarding its content and users in recent years; therefore the need for effective means of accessing and processing available items has attracted a wide range of researchers and industries. Recommender systems has emerged to help both users to find what they may be interested in and the producers to sell their products more efficiently. As the number of these techniques grow, the need to evaluate them properly increases as well. However the proposed evaluation metrics are very diverse and often inconsistent with each other. Although there had been immense research in this field, there is no united and proper approach for evaluation of... 

With the over-increasing growth of the information provided for the web users, from content providing systems and web stores to social networks, the exictence of recommender systems is strongly needed. Recommender systems personalize web for the users and help them with finding relevent information in the huge era of World Wide Web. Collaborative filtering methods are known as the most successful and vastly used recommendation systems. Although they generally outperform content-based algorithms, in cold-start situation and especially in the presence of the new items, they fail to predict ratings for the new items or make good recommendations. This problem is not negligible in the systems... 

Collaborative fltering (CF) is one of the best and widely employed approaches in Recommender systems (RS). This approach tries to fnd some latent features for users and items so it would predict user rates with these features. Early CF methods used matrix factorization to learn users and items latent features. But these methods face cold start as well as sparsity problem. Recent years methods employ side information along with rating matrix to learn users and items latent features. On the other hand, deep learning models show great potential for learning effective representations especially when auxiliary information is sparse. Due to this feature of deep learning, we use deep learning to... 

Study of real networks has become an important research task. Real networks appear in different domains such as social networks and biological networks. It is also possible to extract various networks from the set of scientific collaborations of published papers. In this project we investigate the analysis of citation network and co-authorship networks of scientific papers in order to find out a new method in citation recommendation of scientific papers. Presented method have some superior properties over existing citation recommendation systems like combining different fields for suggesting most relevant papers to user, in response of user’s input keywords. For presenting this method we... 

The recent data explosion phenomena and numerous data consumer services over the world have made recommender systems a crucial component for online platforms.Recommender systems are filtering systems that prune the large volume of information to make the small prefered quota easily and readily accessible to the users. In social networks and news sites, due to the dynamic nature and constantly changing items and user preferences, reinforcement learning in recommender systems has become a hot research topic in the last decade. Our focus of attention,Multi-Armed Bandit, a tool of reinforcement learning, is a sequential decision making problem that illustrates the trade-off between exploring the... 

User cold-start is a common problem among real-world applications in the sequential recommendation field since determining user preference based on a few interactions is difficult. The problem would end up limiting the performance recommender systems. To address the cold-start problem, some previous works used meta-learning along with user’s and item’s side information. Meta-learning algorithms made the model able to share knowledge among all tasks. Although they had promising results, they had some fundamental issues with modeling the dynamics of user preferences and considering all kinds of users’ preferences, especially for minor users. The proposed method includes a model incorporating... 

Nowadays, recommender systems have become powerful tools that engage users in an online manner, over the Internet. Collaborative filtering (CF) is a well established method for building recommender systems and has been applied to several applications. While CF has its advantages,its use is hindered by challenges such as low accuracy for new users (newcomers). With the growth of online social networks, networkbased recommender systems emerged. These systems take advantage of the information available in social networks and the user’s past activity to recognize user behavior and recommend items that are more relevant to each user. One of the most important advantages of network-based... 

Exceptions are an indispensable part of a software development process. Developers usually rely on imprecise results from a web search to resolve the exceptions. More specifically, they manually design a query based on the information indicated by the exception message and the stack trace of that exception. Afterwards, they choose and adapt a solution from the web search results. In this scenario, there is a gap between the development environment and web browsers. Moreover, manual search is very time consuming and a lot of useful information from the code is lost. In this thesis, we introduce Exception Tracer, an Eclipse plugin that helps developers find solutions of exceptions. Exception... 

Mankind with it’s finite resources (Time, Energy, …) cannot make use of every accessible option in daily activities (such as buying items, listening to music and reading news), and is restricted to decide on a handful of them. Available options are increasing on a daily basis and these surplus of available options had an adverse effect; Thus, leading us to more baffling situations. As a result, need for external assistance appeared in decision situations. Considering exceptional computation power available to computers, a framework named Recommender Systems were developed. Recommender systems try to use their accessible data in order to make fitting suggestions to users. Personalization and... 

As the amount of data on the Internet grows, users face diverse options while searching for their desired information and items. Therefore, accessing what one is looking for, is usually time consuming and even impossible in some cases. In order to solve this issue, the goal of recommender systems is to offer recommendations which are compatible with users’ needs and preferences. One of the most important challenges of recommender systems is to improve the quality of recommendations. Recommender systems’ quality can be assessed using different metrics including precision, novelty and coverage. However, these metrics are inconsistent in some applications and improving one will cause a decline... 

With the popularity of smart devices, a lot of applications have developed and deployed.Developers try to establish continuous interaction with their users by different tools including push notifications. Push notification is a message that is sent from developers to the users and as soon as the user’s device receives that, it appears on the device screen. Sending proper content to users in order to resume their engagement is one of the most important usages of notifications. Users are not interested in receiving irrelevant notifications, and receiving irrelevant notifications make them remove the application, so it’s important to predict users’ interest in different notifications and push... 

Smartphones are ubiquitous nowadays. This fact would give the opportunity to the developers to earn lots of money and attention. One of the main factors in a successful business plan is user lifecycle, meaning that users use the app more and more in a specific period. For app developers, notifications are an effective way to interact with the users in different ways depending on the nature of the notification. However, hen a notification is sent out to a user, it is delivered directly without considering the users situationor psychological state. The notification could be perceived as a distraction or interruption, potentially causing inattentionand frustration for the recipient, even if the... 

In this thesis the usage of probabilistic reasoning in collaborative filtering is investigated. The problem of predicting users' rating is formulated as a Bayesian decision problem and a generative probabilistic model is used in order to find the optimal decision. Two different probabilistic models are considered: user based model and rating based model. In user based model prediction of ratings is based on structural learning of Bayesian networks. In rating based model, we assume a predefined Bayesian network represents the joint distribution over model variables and rating prediction is carried out using McMc inference method. MovieLens dataset is chosen to evaluate and compare the results... 

In recent years, most of the programmer’s problems are solved using Q&A websites, e.g. Stack Overflow. Since the number of users, questions and answers in these websites are rapidly increasing, manually moderating and enhancing the quality of the posts (specially the answers) seems to be impossible. Despite the various studies on issues with Q&A websites and facilitating their use, there has been no studies on the field of improving the quality of answers. In this study, we intend to provide an appropriate solution to enhance the problem of low-quality answers. In order to demonstrate the proposed approach, we recommend a tool called StImp, which is a plugin for Eclipse. Using this plugin,... 

The ever increasing growth of smart phones and their applications have led to the emergence of various platforms, i.e. online stores, for distributing them among users and providing a way for customers and developers to interact with each other. In recent years, these stores have turned into an important form of software repositories, which brought out new challenges for researchers. One of these challenges is the task of searching for suitable applications and providing simple access to them for users. Pervious work has focused on the relationships between users and applications’ contents. Among the shortcomings of these approaches, is not considering users’ reviews of applications and what...