Pages: 35

In this chapter an in-depth comparison between Relational Databases (RD) and NoSQL Databases is performed. There is a recent trend for the IT community and enterprises to increasingly rely on the NoSQL Databases. This chapter briefly introduces the main types of NoSQL Databases. It also investigates this trend by discussing the disadvantages of RD and the benefits of NoSQL Databases. The interchanges of the popularity in the past 10 years of both types of Databases is also depicted. Finally, the most important Graph Databases are discussed and the most popular one, namely Neo4j, is analyzed.

Pages: 20

In recent years, we have witnessed an excessive increase in the amounts of data available on the Web. These data originate mostly from social media applications or social networks and thus they are highly connected. Graph databases are capable of managing these data successfully since they are particularly designed for storing, retrieving, and searching data that is rich in relationships. This chapter aims to provide a detailed literature review of the existing graph databases and software libraries suitable for performing common social network analytic tasks. In addition, a classification of these graph technologies is also proposed, taking into consideration (i) the provided means of storing, importing, exporting, and querying data, (ii) the available algorithms, (iii) the ability to deal with big social graphs, and (iv) the CPU and memory usage of each one of the reported technologies. To compare and evaluate the different graph technologies, we conduct experiments related to the link prediction problem on datasets with diverse amounts of data.

Pages: 23

Neo4j is the leading NoSQL graph database. In the past years it has gained constant popularity with ever more businesses and enterprises using it. In addition, many universities have incorporated teaching Neo4j within their syllabi. At the same time, many organizations were forced to change and trust NoSQL databases due to the characteristics of relational databases, such as cumbersomeness and inability to adapt to the needs of data volume eruption and the incorporation of new and diverse data types. This chapter provides a literature survey and analyzes major use cases and applications of Neo4j, focusing on the domain of social media. At the same time, it categorizes them according to identified context. The categorization includes topic detection/extraction, recommendation systems, branding/marketing, learning environments, healthcare analytics, influence detection, fake/controversial information, information modeling, environment/disaster management, profiling/criminality, attractions/tourism, and metrics/altmetrics. For each category, representative examples are showcased elevating the necessity and benefits of using a NoSQL database such as Neo4j for various applications.

Pages: 19

Internet users use social networks for different purposes and with different data. Having the same user accounts in different social networks and combining users’ data in a single graph will improve the functioning of recommendation systems and increase the user experience. In this study, the data of thousands of users in nine different social networks were collected and combined in a single graph. Anchors are created between nodes in graphs with different attributes using the previously proposed node alignment and node similarity methods. Node similarity methods have been developed for multiple social networks, and node matching success has been increased. Thus, it has been possible to propose much more successful recommendation systems. In addition, a new alignment method for multiple social networks is proposed in this study. As a result of the study, the success rates of the proposed methods were measured with the actual data collected from social networks. Graphs from multiple social networks have been converted as a single graph. A broad user profile covering more than one social network has been created for users.

Pages: 20

Influencers gain a growing impact on the social life and economic decisions of the web population. A special group are child influencers, carrying responsibility and popularity unusual for their young age. However, little research data exists regarding child influencers and their follower communities. Furthermore, social networks continue to restrict independent data collection, making the collection of data sets for analysis difficult. This chapter presents an exemplary approach starting from collection of a larger amount of data from YouTube with respect to the access restrictions. By creation of automatic scripts, targeting child influencers and communities on YouTube, followed by storage on the graph database ArangoDB. Through analyzing the data with overlapping community detection algorithms and centralities for a better understanding of the social and economic impact of child influencers in different cultures. With the authors open source WebOCD framework, community detection revealed that each family channels and single child influencer channels form big communities, while there is a divide between the two. The network collected during the research contains 72,577 channels and 2,025,879 edges with 388 confirmed child influencers. The collection scripts, the software and data set in the database are available freely for further use in education and research.

Pages: 24

The smart city paradigm is about a better life for citizens based on advancements in information and communication technologies, like the Internet of Things, social media, and big data mining. A smart city is a linked system with a high degree of complexity that produces a vast amount of data, carrying a very large number of connections. Graph databases yield new opportunities for efficient organization and management of such complex networks. They introduce advantages in performance, flexibility, and agility in contrast to traditional, relational databases. Managing smart city linked data with graph databases is a fast emerging and dynamic topic, highlighting the need for an integrative literature review. This work aims to review, critique, and synthesize research attempts integrating the concepts of smart cities, social media, and graph databases. The insights gained through a detailed and critical review of the related work show that graph databases are suitable for all layers of smart city applications. These relate to social systems including people, commerce, culture, and policies, posing as user-generated content (discussions and topics) in social media. Graph databases are an efficient tool for managing the high density and interconnectivity that characterizes smart cities.

Pages: 15

The smart city (SC) approach aims to enhance populations’ lives through developments in knowledge and connectivity systems such as traffic congestion management, Energy Management Systems (EMS), Internet of Things or social media. A SC is a very sophisticated connected system that generates a big quantity of data and demands a large number of connections. Graph databases (GDB) provide new possibilities for organizing and managing such complicated networks. Machine learning (ML) is positively influencing headline-grabbing apps like self-driving vehicles and virtual assistants, but it is also improving performance and lowering costs for everyday tasks like web chat and customer service, recommendations systems, fraud detection, and energy forecasting. Most of these ML data challenges may be addressed by utilizing GDB. Because graphs are founded on the concept of linking and traversing connections, they are an obvious option for data integration. Graphs can also be used to supplement raw data. Each column in typical tabular data represents one "feature" that the ML system may exploit. Each form of link is an extra feature in a graph. Furthermore, simple graph structures such as causal chains, loops, and forks can be regarded features in and of themselves. Also, Power system engineers have examined several parallel computing technologies based on relational database structure to increase the processing efficiency of EMS applications but have yet to produce sufficiently rapid results. Graph processing does this. Considering Neo4j as the leading NoSQL GDB, the goal of this work is to create and test a method for energy load forecasting (ELF) combining ML and GDB. It integrates multiple approaches for executing ELF tests on historical building data. The experiment produces data resolution for 15 minutes as one step ahead, while disclosing accuracy issues.

Pages: 21

The era of big data constantly introduces promising profits for the healthcare industry, but simultaneously raises challenging problems in data processing and management. The flood of automated systems in our smart cities generates large volumes of data to be integrated for valuable information which can also be applied to support in the healthcare context. Medical databases offer the capabilities to address these issues and integrate, manage and analyze the data for gaining deep insights. Especially modern graph databases support the discovery of complex relationships in ever-growing networks of heterogenous medical data. However, traditional relational databases are usually not capable of representing data networks in tables or revealing such relationships, but are still widely used as data management systems. This chapter discusses a methodology for transferring a relational to a graph database by mapping the relational schema to a graph schema. To this end, a relational schema graph is constructed for the relational database and transformed in multiple steps. The approach is demonstrated for the example of a graph-based medical information system using a dashboard on top of a Neo4j database system to visualize, explore and analyse the stored data.