University of Twente Student Theses
Evaluating Recurrent Neural Networks as Solution to User Action Prediction
Boysen, Yannic (2022) Evaluating Recurrent Neural Networks as Solution to User Action Prediction.
Full text not available from this repository.
| Full Text Status: | Access to this publication is restricted |
| Abstract: | The user interface for enterprise software is often complex, as it has to cater to many different types of users and use cases. For that reason, that type of software could greatly benefit from an adaptive user interface that presents predicted useful functions to the user based on their previous interactions. This thesis is tasked with evaluating the use of recurrent neural networks for the prediction of these user action suggestions on the basis of an industry task for a specific enterprise data protection software. Due to the present unavailability of real user data for this software, the machine learning system is tested with a synthetic dataset based on a created data model, as well as a comparable real dataset of webshop interaction histories. The results show that stochastically generating a dataset with no prior information about distributions is not sufficient to evaluate the performance of neural networks for such a task. While training on the real data achieved comparably superior results, it still showed clear limitations that have to be considered when using this model to solve a specific problem. |
| Item Type: | Essay (Master) |
| Clients: | IBM Germany GmbH |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Interaction Technology MSc (60030) |
| Link to this item: | https://purl.utwente.nl/essays/90532 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page