• What we'll do:
-First, Lisa Röttjers (https://be.linkedin.com/in/lisarottjers) will provide some theoretical background on different Machine Learning algorithms such as KNN, logistic regression and random forest, and then use those algorithms to identify risk factors for cervical cancer (https://archive.ics.uci.edu/ml/datasets/Cervical+cancer+%28Risk+Factors%29).
-Next, Jessa Bekker (https://people.cs.kuleuven.be/~jessa.bekker/) will introduce her work in learning from positive and unlabelled data. This is a special type of data the occurs often in practice. Tradition machine learning requires "fully supervised data" which is "positive and negative data", i.e., each example in the dataset has a label that is either positive or negative. In practice however, labels are often not available and the labels that we do have are all positive. Where does this data occur and what makes it challenging? How can we deal with this kind of data?
-Then, Lisa will follow on the previous talk. She will process the cervical cancer dataset to “unlabel” some of the data and see how that affects tool performance.
• What to bring
A laptop with R and R-studio installed, if you wish to follow the hands-on.
• Important to know
Hello R-Ladies! For the months of March and April, we will be hosting our Meetups in the Agora learning center from KU Leuven. The spots for these are limited, and Agora will require us to provide them with a list of attendees, so that non-KU Leuven affiliated ladies can also join. Because of this, we will ask you to confirm your attendance after you register! Looking forward to seeing you there!