Mood Support, an adaptive social-context aware mood support system
|has title::Mood Support, an adaptive social-context aware mood support system|
|Master:||project within::Human Ambience|
|Student name:||student name::Martin Stolk|
|Second supervisor:||Mark Hoogendoorn|
In this thesis, the conceptual design for a smartphone-based mood-supporting application, Mood Support, is presented. By means of an implementation, the proposed first development phase of Mood Support is realized and tested.
Mood Support is designed to explore the feasibility of mood-enhancing applications within teams through novel use of mobile technology. As such, it is designed to give insight in mood development and its relation to (social) context and a user's behavior patterns. Furthermore, it targets mood improvement of its users by suggesting real-time support actions (SA) during their workday. In order to maximize the effectiveness and user's appreciation of these SAs, Mood Support adapts its given advice based on the user's current mood and (social) context features. These features capture the users' physical activity level, which gains the extent of sedentary time, manually entered events such as the moments of coffee-consumption, and their room-level indoor location measured via iBeacon technology. Moreover, Mood Support combines this location data with time-matched data obtained from other users to get an approximation of the user's social activity level, a proven influencing factor on a person's mood.
By adopting an innovative support action model (SAM) based on a collaborative learning strategy, SAs of the expected highest quality are selected and presented to users without the need of extensive per-user training periods. To accomplish this, the SAM uses a customized distance-weighted k-Nearest Neighbor (DWKNN) algorithm. Although being able to generalize trained data over its users, the SAM is able to have bias towards user-specific data, including the feedback on the effect of SAs and the user's opinion on them.
To test the ability to give insights in the mood development of users within a team and review the performance of the proposed SAM, a practical experiment was devised. Within this experiment, Mood Support's smartphone app collected data and provided the users with real-time SAs. An analysis of the data that resulted from this experiment establishes Mood Support as a valuable asset for mood research and proves that context-aware mood-enhancing applications are indeed viable.