Data Science Portfolio Projects

Navigate within individual folders to see files relevant to that project.

• Dublin Bikes Forecast: I use an ML model to predict Dublin Bike usage for the pandemic period as if the pandemic had never happened.

  • Skills: python, data cleaning, data analysis, data visualisation, ML, forecasting, time series analysis, arima

• Fake News Classification: I use bagging and boosting ensemble methods and compare the results in classifying fake and real news articles.

  • Skills: python, data cleaning, data analysis, data visualisation, ML, NLP, TF-IDF, classification, ensemble methods, random forest, gradient boost

• Forecasting Public Transport Usage: I use a linear regression model and an auto arima model and compare their MAE and RMSE results from the training data. With the auto arima model producing marginally better results and the linear model possibly overfitting the arima model was trained on the full dataset and used for the final forecast. The arima model produced what looks like an accurate forecast for the period required.

  • Skills: R, data cleaning, data analysis, data visualisation, linear regression, auto arima, ML, forecasting, time series analysis

• Twitter Sentiment Analysis: I use almost 250,000 tweets under the #StandWithUkraine hashtag and analysed the sentiment associated with those over the course of the first two weeks of the war in Ukraine. A mapping of real-world events onto the sentiment timeline was also done to attempt to provide explanatory reasoning for the shifts in sentiment.

  • Skills: R, data cleaning, data analysis, data visualisation, sentiment analysis, ML, NLP

• Unordered Multinomial Logit: I construct and interpret an unordered multinomial logit looking at the difference in odds of a country having different levels of GDP (increase, decrease, no difference) based on being a democracy and its oil exports.

  • Skills: R, data cleaning, data analysis, data visualisation, multinomial logit