Modern web information systems use machine learning models to provide personalized user services and experiences. However, machine learning models require annotated data for training, and creating annotated data is done through crowdsourcing tasks. The content used in annotation crowdsourcing tasks like medical records and images might contain some private information which can directly or indirectly identify an individual. The name, age, ethnicity, gender, contact details are examples of private information that directly identifies an individual. Indirect private information relates to the cultural, economic, and social factors of an individual. For instance, the visual cues of religious objects or symbols relate to the religious beliefs of an individual. In this thesis, we study how to minimize the amount of private information extracted from images using a hybrid algorithm which combines machine learning models and crowdsourcing. We also demonstrate that the proposed hybrid algorithm reduces the amount of private information exposed from the image and the cost of using the crowd for detecting private information in the image.

Our world population is increasing significantly resulting in a growing energy demand while the earth is running out if natural resources. This expanded energy demand directly affects climate change. Hence, there is an urgent need to move towards urban sustainability and to reduce our energy consumption. This calls upon a behavioral change in energy consumption by the individuals (i.e., citizens). Social comparison, in the form of comparative feedback on energy usage with others, appears to be a more effective approach to stimulate energy conservation and efficiency than temporal self-comparisons. But before we can motivate people to change their energy consumption behavior, we need to have a thorough understanding of which energy-consuming activities they perform and how these are performed. Thus, insights into the individual’s activities related to energy consumption should be gathered at a high-granular level.
Traditional sources of information about energy consumption, such as smart sensor devices and surveys, can be costly to set-up, may lack contextual information, have infrequent updates or are not publicly accessible. In this research, we propose to use user-generated content - and specifically, social media content - as a complementary source of information due to its rich and semantic nature. A huge amount of social media data is generated by hundreds of millions of people every day. These data sources are also publicly available and provide real-time data which is often tagged to space and time. Social media data also contains a lot of meta data, making it a good source for the recognition of energy-consuming activities performed by individuals.
This thesis contributes the Social Smart Meter framework in order to meet the aim of automatically processing user-generated content for the description of energy-consuming activities, both at individual and group level. Four different categories of energy-consuming activities are distinguished: dwelling, food consumption, leisure, and mobility. To get a better understanding of the domain of energy-consuming activities, we contribute the Social Smart Meter Ontology (SSMO). This ontology forms the base for the data processing pipeline, which is developed in order to collect and enrich the data using several state-of-the-art techniques. Hereafter, the enriched data is classified to the different categories of energy-consuming activities using a dictionary- and rule-based approach, along with a classification confidence. To find ground truth and to evaluate the framework’s performance, a user-based evaluation approach was used.
Furthermore, we contribute a Web-based application to support the analyses at group (i.e., city and neighborhood) level. Case studies are performed for the cities of Amsterdam and Istanbul, for which 275K social media posts are collected. The aggregated results are analyzed, providing more insights into the energy-consuming activities identified in the collected social media content. The majority of the classified social media posts refers to leisure activities. In addition, by examining for each post whether there exists a (significant) distance to the previous post created by this user, many mobility activities are inferred.
The case studies also contribute to the evaluation and discussion of the framework’s performance; by analyzing its results, the framework’s adherence to reality was discussed. Based on our preliminary results, it seems that using user-generated content has great potential as a complementary source of information for identifying and describing energy-consuming activities that are not yet captured by traditional data sources.

Global trends such as digitalization and the Internet of Things have influenced cities and the lives of residents deeply in recent years. In this digital age, cities turn toward SMART solutions, where they ultimately hope to find new innovative ways of accommodating their citizen’s fundamental needs such as clean drinking water, efficient sewage systems, road traffic and logistical systems or fire and flood control. On the one hand digitalization offers so far unexplored opportunities. On the other hand new issues, such as cyber security and privacy need to be addressed. Apart from these new developments, there are long-standing challenges and the question of who the city belongs to, which are furthering the prevailing inequality in cities and metropolitan areas. These issues are reflected in socio-economic segregation, meaning the widening gap between rich and poor, and the development of ‘concentration areas’, which are the result of economic disadvantage and social inequality.
This research aims at unravelling the benefits of the digital age with regard to managing and enhancing urban dynamics on a very local scale. Here, big data offers the opportunity of retrieving real-time data with regard to the city and its residents, on a detailed scale, which enables the study of intrinsic social structures of neighbourhoods. Information on the complexity of social urban dynamics is essential in decision making processes of urban developments and construction projects. Especially since redevelopment projects have become the answer to the long-standing challenges cities are faced with.
The scope of the following study has been framed within the context of redevelopment projects in Amsterdam, observing the outcome of so-called ‘state-led gentrification projects’ and exploring the opportunities big data holds to function as a decision making support in managing and enhancing urban development and construction processes of the built environment.

‘Real estate is the largest asset class in the world’ as stated by Harvard business school Professor Arthur Segel. It is also unbeatable driver of individual wealth and overall economy. But on the other hand, it is one of the most imperfect markets, due to the lack of accessibility to valuable information in the limited time of decision making process. Thus, it is crucial to find out what is brought by the state of art that could tackle this problem in such a complex and competitive context.

Since, the main component of information is data which is also the chief resource in the modern world; (big) data could be a hot phenomenon to be adapted within the real estate investment domain for improving the performance of the real estate market.

Thus, the main goal of this research is to bridge the area between real estate investment domain and big data, by leveraging big data methods, predictive analytics and smart tools for achieving informed real estate investment decisions. Consequently, investors will be able to maximize return, achieve better risk diversification and select the right time to invest.

The expected final product of this research is a developed integrated decision making model and flow chart that involves big data methods and techniques for making more informed real estate investment decisions ‘Dutch office market’.