Here is a selection of projects I have recently completed. Feel free to have a look around! Alternatively, the portfolio is also available to view on my website.
-
Machine Learning projects:
-
Credit card fraud: improving detection with Deep Learning: comparing different models performance at predicting instances of card fraud. The final model is a 5-layer Neural Network built with Keras that correctly classifies 99.96% of transactions.
-
Creating customer segments using different clustering algorithms: unsupervised learning project in which we create clusters of customers of a wholesale distributor using different clustering algorithms, including K-means and DBSCAN.
-
Predicting Boston housing prices: using a Machine Learning model to predict house prices in the Boston metropolitan area. Identifying the optimal price for clients wishing to sell their home.
-
Finding donors for CharityML: in this project, we compare several Machine Mearning algorithms in order to accurately model individuals' income and predict who is likely to become a donor.
-
Mushrooms: safe to eat or deadly?: predicting whether mushrooms are edible or poisonous using different Machine Mearning models.
-
Exploratory analysis: Titanic dataset: using Seaborn visualisations, this project explores the passengers' details and reveals which factors helped some of them survive the shipwreck.
-
-
Natural Language Processing projects:
-
Sentiment analysis: US airlines on Twitter: using a pre-trained GloVe embedding, we compare how well a bidirectional GRU and a birectional LSTM predict the sentiment of tweets regarding US airlines.
-
Detecting spam text messages: predicting whether a text message is spam or legitimate with different models. The best performance is obtained with a Multinomial Naive Bayes model and a CountVectorizer approach.
-
Toolkit: Python, Keras, Scikit-Learn, Pandas, NLTK, TensorFlow, Matplotlib, Seaborn