direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Es gibt keine deutsche Übersetzung dieser Webseite.

Master Thesis: Design and Implementation of a Mobile Platform for Recommender Systems


Design and Implementation of a Mobile Platform for Recommender Systems


In recent years, recommender systems (e.g., Spotify, Netflix or Amazon) have evolved and become ubiquitous in everyday life. However, the aforementioned platforms are mostly limited to only recommend items that the platform itself offers. Moreover, recommendations by these platforms might not necessarily be based on the user’s preferences as providers might, for example, push items or products that create the biggest margin. Besides, platform providers often try to create so-called lock-in effects that aim to force users to stay with the current provider. In this respect, the afore given examples are centralized – one service provider controls all the data and determines how to calculate recommendations. In this thesis, I present an alternative solution for a recommender system platform that is not limited to one single service provider. The modern smartphone that already stores lots of relevant data about its owner (e.g. favorite music) combined with its capabilities for device-to-device communication provides the technological infrastructure for such a platform. Users can exchange data with other nearby users and create local databases on their devices that serve as a basis to calculate recommendations. I present a mobile architecture that is built in a modular way and consists of three main components: Data Collection, Data Exchange and Recommender System. Although, technologies for short-range communication are widely available in current smartphones, there is still no fully developed solution to exchange larger messages between Android and iOS devices. Additionally, exchanging messages in the background still poses an obstacle. Therefore, I developed a work-around that integrates the Google Nearby Messages API to broadcast messages that contain a file URL that references to a public file in a cloud storage that stores the actual data. To determine similarity between users, I compared the user’s location traces using a Counting Bloom filter. I implemented the proposed architecture in the form of a mobile prototype for Android and iOS. The implementation works cross-platform and was developed using the Ionic Framework, which is based on Angular and Apache Cordova. An evaluation of the prototype showed that the data exchange works reliably across 100 meters and messages from other devices are received reliably within 4 minutes (new users) and 60 minutes (previously met users). Future improvements are necessary with respect to the battery drainage which is extremely high with 5.29%/h on Android and 10.51%/h on iOS.

Supervisor: Felix Beierle

Type:  Master Thesis

Duration: 6 months

Zusatzinformationen / Extras


Schnellnavigation zur Seite über Nummerneingabe

TU Berlin - Service-centric Networking - TEL 19
Ernst-Reuter-Platz 7
10587 Berlin, Germany
Phone: +49 30 8353 58811
Fax: +49 30 8353 58409