Development of a model to predict outcomes after dynamic cycling people with Parkinson's disease
Our models strive to provide the first steps towards teasing out what variable might contribute to participant response to the dynamic bike, with the intention to use these models in a future adaptive bike that will adapt to participant's unique characteristics to improve outcomes. Current suggestions include a focus on maximizing participant effort and conduction of repeated measures studies.
The problem: members in our lab do not have any programming experience.
The solution: I created an intuitive webgui through a combination of conda, batch scripting, and the Streamlit python library. The script guides users through the 4 steps to obtain the sample, approximate, and spectral entropies of time series outputs from our dynamic bike with just a simple double click on the batch script.
The problem: friends were having a hard time dividing a DoorDash/UberEats bill, making sure everyone paid for only their own food items + tax/tip/fees
The solution: a series of scripts presented with Streamlit python library that takes as input the item costs per person, taxes, fees, and presents the user with customized Venmo links
The problem: media report covid numbers in different ways, the major dashboards are full of information to the point that it's hard to keep track at a glance
The solution: a dashboard made with the Streamlit library that automatically downloads the latest case and vaccination numbers for each state in the US and each country in the world, presenting them with Plotly.
The problem: the dynamic bike v2 file outputs have to be manually reorganized and cleaned prior to entropy analysis. This is a time consuming process.
The solution: this software reorganizes the files created by our lab's stationary bike, gives the option to clean the data and obtain basic statistics like the mean and standard deviations. The newly reorganized files are ready for entropy analysis in MatLab.
The problem: our lab collected dance movement data but it was not compatible with our engineer’s MatLab script for calculating entropy
The solution: a series of python scripts that formats raw output files into one our engineer’s MatLab script can understand, calculates the lowest autocorrelational function’s lowest lag value, and feeds this information into the MatLab entropy script
The problem: mid sized streamers are having trouble with selecting optimal clips from their Twitch streams for their story telling
The solution: feature engineering to create appropriate predictors of the “best” clip, then using this dataset to train and test a categorization machine learning model to provide users with recommended clips
The problem: data collected from accelerometers attached to participants are not compatible with the engineer’s MatLab script for calculating entropy
The solution: a series of scripts to clean raw data with R and calculate the entropy of movement using various python libraries. The results were presented at multiple conferences and published in the Medicine and Science in Sports and Exercise journal.