Women Who Code Portland and Portland R User Group are coming together to host a night full of talks on R and Machine Learning and an opportunity to network with like-minded individuals.

Program

6.00 - 6:30: Doors Open

6.30 - 7:30: Talks

7:30 - 8:00: Networking and wrap up.

Talk #1 - Predictive Analytics and Machine Learning in R - Myffy Hopkins

All data has a story to tell. You can force a story upon it, or you can breath life into it to let it speak for itself. Myfanwy "Myffy" Hopkins has been using advanced statistical methods, R, and machine learning for over 10 years to transform data into information for decision making. The tools provided by R packages are unparalleled at quickly transforming data into a working predictive model. Myffy brings the statistical and data mining expertise needed to make R a highly productive space for generating predictive models using machine learning methods. In her talk, Myffy will show how R makes the difficult statistical concepts of Predictive Analytics manageable to execute. Seeing the actual results of a predictive model will help get through the more grueling parts of the learning curve to becoming a Data Scientist.

Talk #2 - Automated Feature Selection of Predictors in Electronic Medical Records Data - Jessica Minnier

Jessica Minnier is an assistant professor of biostatistics at Oregon Health & Sciences University in the OHSU-PSU School of Public Health. Her research interests include statistical methods for risk prediction and classification, the analysis of 'omics data, and the analysis of large data sets such as electronic health records data. She is interested in reproducible research as well as statistical computing and data visualization with R and shiny.

In her talk, Jessica will talk about "emrselect" - https://github.com/jminnier/emrselect. "emrselect" is an R package that automates the feature selection method for phenotype prediction with Electronic Medical Record (EMR) data to reduce the number of candidate predictors and in turn improve model performance by relying entirely on unlabeled observations. The proposed method generates a comprehensive surrogate for the underlying phenotype with an unsupervised clustering of disease status based on several highly predictive features such as diagnosis codes and mentions of the disease in text fields available in the entire set of EMR data. A sparse regression model is then built with the estimated outcomes and remaining covariates to identify those features most informative of the phenotype of interest. Empirical results suggest that this procedure reduces the number of gold-standard labels necessary for phenotyping, thereby harnessing the automated power of EMR data and improving efficiency.

Who Should Attend?

Anyone who is interested in exploring the world of R and Machine Learning or just curious to learn what they are. By attending our event, you are agreeing to support our mission and follow our Code of Conduct.