As Machine Learning is becoming more accessible to small businesses, thanks to the rapid advance in computing power, smaller start-ups such as Sjauf (a ride sharing start-up) are starting to get interested in implementing Machine Learning solutions in their product. Sjauf needed a system that could automatically tell its customers how much a certain trip would cost them. Using this information multiple different models were developed and integrated into an ensemble. This ensemble as well as the models used by it were then used for price prediction. This project is a proof of concept to show that Machine Learning is capable of solving this problem in real time.

After researching state of the art Machine Learning models for price recommendation, the architecture of the system was designed. The supplied data was preprocessed, after which a custom Genetic Algorithm was developed for optimising models and ensembles. After validation on real-life company data, a comparison using empirical metrics was conducted. We use these empirical metrics to show that a bagging ensemble is the most efficient and accurate model for this purpose. This bagging ensemble outperformed the currently implemented functions, whilst adhering to the set boundaries on response times. Lastly, recommendations are made to the company with an overview of potential future work in this subject.
... Read more

Social media platforms are used by a massive, growing number of users, who use these platforms to share content such as text, photos, videos, and location information. As the spread of social media is playing an increasingly important role in our world, literature has shown that while aiming to promote a number of human values (e.g. friendship, social recognition, and safety), this type of technology may pose risk to other values (e.g. privacy and independence), creating what has been defined as value tensions. This thesis proposes the norm-based, Social Commitment (SC) models as a solution that could potentially provide tailored support for user values. As research shows that norms can fulfill (or pose risks to) values, SC models could utilize their normative core to as well as their ability to contain key relevant information that complement the missing features in social applications’ preference settings, to give users a rich, flexible, and adaptive structure that improves their social application experience. Location sharing in the family life (i.e. within families with children in the elementary school age) was selected as an application domain, as it provided potential use cases that are abundant with value tensions (e.g. a child’s safety vs. their independence), while embodying the essential elements of data sharing using social platforms. The research followed a Situated Cognitive Engineering approach, and an exploratory investigation into the social context of the application domain was conducted: focus groups and cultural probing studies with parents and children, and the collected data was analyzed using grounded theory. The result was a grounded model that showed (1) how activities, concerns, and limitations related to family life are connected through specific user values, and (2) that norms can support these values by promoting activities, alleviating concerns and overcoming limitations. Further on, a conceptual model was built, and subsequently a SC grammar (and a semantic lifecycle) were developed for this domain: the SC-model allowed users to construct commitments amongst each other for sharing and receiving social data, harmonized to their values via normative statements. A location-sharing application was developed so that, in addition to location-sharing features found in familiar commercial platforms, it also contained an implementation of our SC grammar. The SC model’s expressivity was validated through a qualitative user study with parents and children, where nearly all participants’ normative statements were found to be expressible through the proposed model. The SC grammar’s usefulness (within the application domain) as well as its ease of use were validated through a crowd-sourced, online user study. The SC model’s ability to provide improved human value support was validated through a user study conducted with elementary school children using the location-sharing application we developed, as well as a questionnaire constructed to measure fulfillment of children’s values relevant to the domain. Results demonstrated that enhancing the app with the SC model has improved its support a number of children’s values while posing no risk to the remaining measured values in the process. In the thesis’s final user study, we demonstrated that using contextual information (e.g. a user’s value profile) as well as commitment attributes (e.g. recency and norm type), can be used to create predictive models that are capable of automatically resolving the vast majority of conflicts that may occur amongst location-sharing commitments. In conclusion, this thesis demonstrates that SC models possess the potential to provide an easy to use, flexible tool that allows social applications to work better in users’ favor, ... Read more

Whether we are talking about our research at a conference, making a speech at a friend’s wedding, or presenting a proposal in a businessmeeting,we have to speak in public from time to time. How well we deliver a presentation affects the way people think about us and our message. To deliver a well-received speech, preparation is necessary. Among various speech preparation activities, practicing with an audience is regarded as an effective way for enhancing speech performance. However, it is often impractical to organize an audience to practice a presentation and to arrange the diverse set of audience behaviours that are tailored to trainee’s individual skills and learning goals. Virtual reality can provide a solution by practicing with a virtual audience. Although virtual audiences have been used in many domains, e.g., evoking social stress, therapy for social phobia, and improving teaching performance, little research has been reported on the impact of virtual audiences on public speakers’ belief and performance. Therefore, this thesis aims to create a virtual audience which generates flexible expressive behaviours for a public speaking scenario and examines how public speaking experiences in front of such an audience affect the speakers’ belief and speech performance.
... Read more