Himel Dev

A user interaction based community detection algorithm for online social networks

Existing community detection techniques either rely on content analysis or only consider the underlying structure of the social network graph, while identifying communities in online social networks (OSNs). As a result, these approaches fail to identify active communities, i.e., communities based on actual interactions rather than mere friendship. To alleviate the limitations of existing approaches, we propose a novel solution of community detection in OSNs.

User Interaction Based Community Detection in Online Social Networks

Discovering meaningful communities based on the interactions of different people in online social networks (OSNs) is an active research topic in recent years. However, existing interaction based community detection techniques either rely on the content analysis or only consider underlying structure of the social network graph, while identifying communities in OSNs. As a result, these approaches fail to identify active communities, i.e., communities based on actual interactions rather than mere friendship. To alleviate the limitations of existing approaches, we propose a novel solution of community detection in OSNs. The key idea of our approach comes from the following observations: (i) the degree of interaction between each pair of users can widely vary, which we term as the strength of ties, and (ii) for each pair of users, the interactions with mutual friends, which we term the group behavior, play an important role to determine their belongingness to the same community. Based on these two observations, we propose an efficient solution to detect communities in OSNs. The detailed experimental study shows that our proposed algorithm significantly outperforms state-of-the-art techniques for both real and synthetic datasets

A hierarchical approach for identifying user activity patterns from mobile phone call detail records

With the increasing use of mobile devices, now it is possible to collect different data about the day-to-day activities of personal life of the user. Call Detail Record (CDR) is the available dataset at large-scale, as they are already constantly collected by the mobile operator mostly for billing purpose. By examining this data it is possible to analyze the activities of the people in urban areas and discover the human behavioral patterns of their daily life. These datasets can be used for many applications that vary from urban and transportation planning to predictive analytics of human behavior. In our research work, we have proposed a hierarchical analytical model where this CDR Dataset is used to find facts on the daily life activities of urban users in multiple layers. In our model, only the raw CDR data are used as the input in the initial layer and the outputs from each consecutive layer is used as new input combined with the original CDR data in the next layers to find more detailed facts, e.g., traffic density in different areas in working days and holidays. So, the output in each layer is dependent on the results of the previous layers. This model utilized the CDR Dataset of one month collected from the Dhaka city, which is one of the most densely populated cities of the world. So, our main focus of this research work is to explore the usability of these types of dataset for innovative applications, such as urban planning, traffic monitoring and prediction, in a fashion more appropriate for densely populated areas of developing countries.

Identifying Frequent User Tasks from Application Logs

In the light of continuous growth in log analytics, application logs remain a valuable source to understand and analyze patterns in user behavior. Today, almost every major software company employs analysts to reveal user insights from log data. To understand the tasks and challenges of the analysts, we conducted a background study with a group of analysts from a major software company. A fundamental analytics objective that we recognized through this study involves identifying frequent user tasks from application logs. More specifically, analysts are interested in identifying operation groups that represent meaningful tasks performed by many users inside applications. This is challenging, primarily because of the nature of modern application logs, which are long, noisy and consist of events from high-cardinality set. In this paper, we address these challenges to design a novel frequent pattern ranking technique that extracts frequent user tasks from application logs. Our experimental study shows that our proposed technique significantly outperforms state of the art for real-world data.

AntiqueData: A Proxy to Maintain Computational Transparency in Cloud

Cloud computing offers computing and software services to users on an on-demand basis. It facilitates users to use computing resources as utility with pay-per-usage billing, which allows users to acquire computational resources with low or no initial cost. Due to this greater level of flexibility, the cloud has become the breeding ground of a new generation of products and services. Since more and more people rely on the cloud with their data and computing, ensuring the trustworthiness of cloud services has become a major issue for both the users and cloud providers. Due to the black box nature of cloud, there has been a lack of trust among providers and users, which has become a major barrier to the widespread growth of cloud computing. One of the trust concerns of cloud is lack of computational transparency. In particular, in current cloud architecture a provider controls all the logging and auditing records corresponding to computation and users do not have access to these records. This is a big concern for many clients of cloud. In this paper, we first identify the risks associated with lack of transparency in cloud and propose a middleware service that eliminates these risks.

A Real-Time Crowd-Powered Testbed for Content Assessment of Potential Social Media Posts

Increasing eminence of online reputation of individuals along with the tendency to avoid unpleasant real-life and/or virtual events such as cyber-bullying, social awkwardness, unintentional false news propagation etc., have made many social users concerned about their social media posts. To avoid the miscellaneous unpleasant events and to ameliorate online reputation, rigorous assessment of proposed posts before broadcasting them in actual social media has become crucial for these users. We observe that, such pre-screening of a proposed post requires human evaluation or feedback regarding different aspects of the post, which in turn assists the associated user in deciding whether or not he/she should broadcast the post in actual social media. In this paper, we address this issue and propose a crowd-powered testbed that allows a social media user to get a real-time evaluation of his/her proposed post. This assessment of a proposed post includes a positive/negative recommendation indicating whether or not the post should be broadcasted in actual social media.

Privacy preserving social graphs for high precision community detection

Discovering communities from a social network requires publishing the social networks data. However, community detection from raw data of a social network may reveal many sensitive information of the involved parties, e.g., how much a user is involved in which communities. An individual may not want to reveal such sensitive information. To resolve this issue, we address the problem of privacy preserving community detection in social networks. More specifically, we want to ensure that community detection is possible from the published social graph/data but the identity of users involved in a community should not be disclosed.

The Size Conundrum: Why Online Knowledge Markets Can Fail at Scale

In this paper, we interpret the community question answering websites on the StackExchange platform as knowledge markets, and analyze how and why these markets can fail at scale. A knowledge market framing allows site operators to reason about market failures, and to design policies to prevent them. Our goal is to provide insights on large-scale knowledge market failures through an interpretable model. We explore a set of interpretable economic production models on a large empirical dataset to analyze the dynamics of content generation in knowledge markets. Amongst these, the Cobb-Douglas model best explains empirical data and provides an intuitive explanation for content generation through the concepts of elasticity and diminishing returns. Content generation depends on user participation and also on how specific types of content (e.g. answers) depends on other types (e.g. questions). We show that these factors of content generation have constant elasticity and a percentage increase in any of the inputs leads to a constant percentage increase in the output. Furthermore, markets exhibit diminishing returns-the marginal output decreases as the input is incrementally increased. Knowledge markets also vary on their returns to scale-the increase in output resulting from a proportionate increase in all inputs. Importantly, many knowledge markets exhibit diseconomies of scale-measures of market health (e.g., the percentage of questions with an accepted answer) decrease as a function of the number of participants. The implications of our work are two-fold: site operators ought to design incentives as a function of system size (number of participants); the market lens should shed insight into complex dependencies amongst different content types and participant actions in general social networks.In this paper, we model the community question answering (CQA) websites on Stack Exchange platform as knowledge markets, and analyze how and why these markets can fail at scale. Analyzing CQA websites as markets allows site operators to reason about the failures in knowledge markets, and design policies to prevent these failures. Our main contribution is to provide insight on knowledge market failures. We explore a set of interpretable economic production models to capture content generation dynamics in knowledge markets. The best performing of these, well-known in economic literature as Cobb-Douglas equation, provides an intuitive explanation for content generation in the knowledge markets. Specifically, it shows that (1) factors of content generation such as user participation and content dependency have constant elasticity--a percentage increase in any of the inputs leads to a constant percentage increase in the output, (2) in many markets, factors exhibit diminishing returns--the incremental, marginal output decreases as the input is incrementally increased, (3) markets vary according to their returns to scale--the increase in output resulting from a proportionate increase in all inputs, and finally (4) many markets exhibit diseconomies of scale--measures of market health decrease as a function of overall system size (number of participants)

Cassandra: a crowdsourced testbed for content assessment of potential social media posts

Content assessment of posts before broadcasting them in social media has become crucial for many social media users. Primary reasons include online reputation management, avoiding awkwardness in social media, preventing cyber-bullying, preventing unintentional false news propagation. We observe that, such content assessment of a proposed post requires human evaluation or feedback regarding different aspects of the post in order to assist the associated user in deciding whether or not s/he should broadcast the post in social media. In this paper, we address this issue and propose a crowdsourced testbed that allows a social media user to get an evaluation of his/her proposed post before broadcasting it in actual social media, based on the feedback of specialists associated with the topics of the post. This assessment of a proposed post includes a positive/negative recommendation indicating whether or not the post should be broadcasted in social media.