Machine Learning
Below is a list of different problems that were solved with Machine Learning in Python and in R. The concepts that are covered range from Network Analysis to Optimization as well as Time Series.
Predicting Cryptocurrency Value Using Twitter
The goal in this project was to predict the price fluctuations of a cryptocurrency called Cardano. This was made possible by measuring the sentiment score of the crypto using Twitter and a Python package called VADER. This technique allows us to properly foresee 73% of the changes to the value of a coin.
Evolution of COVID-19 in France
A research on the effect of COVID-19 in France was done using R as the main coding language. Visuals of the different Coronavirus metrics were created, those metrics being the number of deaths and the number of active cases for theoretical situations where only a part of the population is vaccinated or where the whole population gets the jab. Doing so, the effects of government restrictions such as mandatory quarantine or the curfew can be measured in numbers.
Network Analysis of a Super Bowl Ad Campaign
A Network Analysis of General Motors’ Twitter presence was done after their 2021 Super Bowl Ad Campaign. Different buzz words as well as major influencers were isolated to get a better sense of the company’s different centrality measures. That marketing campaign led to a sharp increase in how positively Twitter users viewed GM as a brand.
MLFlow to Predict Bankruptcy
Used a little under 100 predictors to predict whether a company goes bankrupt within the next 5 years. This task was accomplished with the use of MLflow to optimize the parameters of a LightGBM model. This optimized model gave us 90% accuracy in finding unsuccessful businesses.