GCHESTER.COM

This Jupyter notebook tool was developed to enable the visualisation of similarities and differences of the major texts from many of the worlds largest religions. It also includes non-religious works from approximately similar periods (of translation) as a benchmark. While the tool is focused on religious works, it is general enough to be applied to be used to visualize the comparison of books from any genre.

Finds the best vocabulary of words to determine sentiment of movie reviews (using t-test). Then uses this vocabularly and logisitc regression on TFIDF word vectors to predict sentiment on 3 test datasets. Compare performance using AUC of ROC curve.

This Jupyter notebook performs various data transformations, and applies various machine learning classifiers from scikit-learn (and XGBoost) to the a loans dataset as used in a Kaggle competition. It also uses scikit-opt Bayesian optimisation to find the best hyperparameters.

Forecasting Walmart Store and department sales, using seasonality/trend decomposition in R

This Jupyter notebook performs various data transformations, and applies various machine learning algorithms from scikit-learn (and XGBoost) to the Ames house price dataset as used in a Kaggle competition.

Health Data was sourced from the CDC 500 cities prokect and combined with life expectancy data from healthinequality.org. Javascript and D3 were used to show multiple views of this data, enabling detailed city by city comparison of health and life expectancy.

Various Tensorflow Models based on the tensorflow-provided tutorials but with improved performance

Health vs Wealth Academic Paper - CCA Big Data project

This notebook uses Spark (via PySpark) to read HDFS files from an AWS S3 bucket, and to join them and perform some basic analysis (correlations). Pandas is used to improve output formattting.

How to use tSNE for visualisation of high-dimensional data (Jupyter notebook)

Python Jupyter notebook - lots of EDA then XGBoost for modelling.

Jupyter notebook in Python- some EDA and lots of scikit-learn modelling

Collection of Useful Python Data Science Cheat Sheets - Pandas, Numpy, Matplotlib, Scipy, Scikit-Learn, PySpark, and Keras

Experiments with PYMC3, including finding mean and std, linear regression adnd solving the German Tank Problem

How to install Theano Python Neural Network package on Windows 10 with Python 3.5

Kaggle Titanic competition - SVM and Random Forest entries. The key to good results was creating the right features and then tuning the classifiers, then back to the features and finally a re-tune of the classifiers. Arguably the classifiers are too finely tuned and a 'real' result should be about 1% less than that submitted.

How to setup a blog on GitHub with Pelican - Part 2

How to setup a blog on GitHub with Pelican

Online and in-person training

Attempted Summary of the Python Ecosystem

Useful books for Python and Data Science

Stock price tinkering, very early attempt. Uses quandl to download stock prices from google finance.