Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Improving the efficiency and enhancing the capacity of the PyPLT (Python Preference Learning Toolbox) software tool

Chaviara Antonia-Chrysovalanto

Full record

---

Summary

Research has demonstrated that ordinal approaches to the analysis of subjective values such as emotions, leads to more reliable predictive models. Preference learning is the machine learning subfield, which deals with datasets including ordinal relations. Preference learning algorithms have proven to be powerful in creating efficient computational models from ordinal data. The Python Preference Learning Toolbox facilitates ordinal data processing and preference learning. The software is open source, available to a wide range of researchers and includes popular algorithms and data processing methods. At first, the toolbox is tested with synthetic datasets in order to identify possible malfunctions during the stages of data processing and modelling. An optimization of the current features, along with the addition of evaluation metrics and preference learning techniques are performed in order to augment the functionality of the software. A user survey follows, in order to test the usability of the toolbox. The results confirm that PyPLT is simple, easy to use, for both novice and experienced researchers. Furthermore, it is capable of producing reliable predictive models provided the necessary data processing and algorithm parameterization which is offered by the toolbox.

Available Files

Services