Recommending healthy bundles in a food context

Studenteropgave: Speciale (inkl. HD afgangsprojekt)

  • Kristian Bjerglund Dyhrberg
  • Mark Vinsløv Nielsen
  • Bogi Hansen Eliasen
4. semester, Datalogi (it), Kandidat (Kandidatuddannelse)
This thesis looks at how to recommend
healthy bundles consisting of dinner
recipes. The proposed approach to
evaluate the healthiness of recipes is

based on the proportions of ingredi-
ents. Furthermore, a bundling algo-
rithm from previous work is adjusted

to include the developed health mea-
sure along with a trained SVD model.

Different versions of the algorithm is
tested in an offline setting in order to
evaluate which version performs best.
The accuracy of the SVD model is also
compared with a KNN algorithm to
determine if KNN is a viable option to
SVD. Due to the complexity of bundles
in regards to presentation, a user study
with 16 participants is also conducted.

The user study tests three different lay-
outs with the purpose of finding which

is best perceived.
Udgivelsesdato14 jun. 2018
Antal sider78
ID: 280887033