About

Helix is an application that allows users to train and test machine learning models in a no-code fashion. It enables users to see which features in their dataset have the greatest impact in the predictions their models make. Helix also produces performance metrics for the models and publication ready figures about performance and feature importance. It also leverages fuzzy logic to determine which features have a “low”, “moderate” or “high” impact on predictions. Furthermore, fuzzy logic is used to determine which features have what impact on predicted categories in the case of classification problems. For regression problems, where the prediction is a continuous quantity, like a price, the user can define categories like “cheap” and “expensive” and then see which features have what impact on whether prediction is “cheap” or “expensive” – or whatever categories the user defines. For a more detailed description of the fuzzy logic process, see Rengasamy et al., 2022 (references).

Helix began life as a command-line tool that could be used to train machine learning models, analyse their performance, describe feature importance in the data and use fuzzy logic to elucidate the level of impact each feature has on the models’ predictions. However, the principal investigator behind it, Grazziela Figueredo, wanted a more user-friendly graphical interface (GUI) and also to go through the code and make it more sustainable and extensible. They had already begun a basic GUI using streamlit in Python. DRS came in with their Python, machine learning and software engineering expertise to help build the GUI app following best practices. This included adding automated testing of code for quality and functionality, refactoring code to make it reusable, using the Agile framework for planning and documentation.

Now, Helix is a downloadable application published on the Python Packaging Index (PyPI) and can be installed on machines running Python.

References

Rengasamy, D., Mase, J.M., Kumar, A., Rothwell, B., Torres, M.T., Alexander, M.R., Winkler, D.A. and Figueredo, G.P. (2022). Feature importance in machine learning models: A fuzzy information fusion approach. Neurocomputing, 511, pp.163–174. doi:https://doi.org/10.1016/j.neucom.2022.09.053.

Links

Helix on PyPI: https://pypi.org/project/helix-ai/

Helix on GitHub: https://github.com/Biomaterials-for-Medical-Devices-AI/Helix

Installing Helix: https://biomaterials-for-medical-devices-ai.github.io/Helix/users/installation.html