Another busy week working with scikit-learn! Really starting to come together I think! I was able to continue to iterate on the Breast Cancer Data tutorial we were assigned as well as begin to work on writing some additional methods using different functions in scikit-learn I was exposed to in my independent work. I also did some modeling on the Iris dataset, a machine learning dataset published by R.A. Fischer whose paper “The use of multiple measurements in taxonomic problems” is still referenced today and is considered a classic in the field of pattern recognition. The goal is to be able to differentiate between three possible classes of Iris plant using the features of petal length and width, as well as sepal length and width. This dataset is also included in the scikit-learn api so requires no additional file downloads and can be included in a project with a simple import statement.
I employed the SVM (Support Vector Machine) model for classification of the iris dataset, achieving an accuracy of 93% after training on 80% of the dataset and using the remaining 20% for testing. SVM plots data using an x vector and then attempts to identify decision boundaries(support vectors) in order to draw a hyperplane, or line, which will divide/separate the data using these decision boundaries to support a best fit that maximizes the margin between the support vectors. The goal is that this hyperplane should be created in a way that separates the data as best as possible, reducing the potential for misclassification/mistakes. I also began the implementation of linear and logistic regression on another dataset included in the scikit-learn dataset api, but did not have time to finish those tasks. I hope to finish up working with those models next week and implement some experiments using them on the Breast Cancer dataset.
Diversity Training
This week we also attended Diversity Training, led by Christina Crawford, focused on MicroAggressions in the workplace. I really enjoyed this presentation and how she kept the content very engaging by dividing the attendees into breakout rooms often and allowing time for small group discussion and then bringing us all back together to share the ideas and opinions we discussed and connect them to the core concepts she addressed. The focus of this training largely addressed how to identify a microaggression, what traits make an event or interaction a microaggression, and how to avoid microaggressions through sensitivity and awareness.
We also had a Cardiovascular System and Photoplethysmography talk by Mel Boonya-Ananta from Florida International University. He identified that current methods of taking a patient’s blood pressure are invasive, requiring an arm or wrist cuff and pressure on the area, and non-continuous, or are taken in intervallically rather than being in a state of constant monitoring, which only offers a snapshot of that measure at a specific instance in time. With this in mind, Boonya-Ananta’s work involves creating a model of continuous cardiovascular monitoring with simulated PPG utilizing a wrist worn device. I really enjoyed this presentation, specifically his discussion of FEA (Finite element analysis), a model used to predict how some structure or device would react to real world forces such as pressure or heat.
