TDADS

Technology Degree Apprentice Data Science Project

This is a fork of this repository in its final state for archival purposes.

I worked on this project from June 2020 to August 2020 with other PwC Technology Degree Apprentices, and the project was split into two tasks.

Task 1 (June-July)

Categorisation of digits using the MNIST dataset:

• This was a more defined task, with the goal simply being to create the most accurate model for categorising digits.
• We created various models - both neural networks and non-neural networks - and produced a presentation detailing our findings.

Personally, I worked on three models:

• Support Vector Machine (SVM): Implemented using sklearn
• Naive Bayes: Custom implementation using numpy
• Random Forest: Implemented using sklearn

Task 2 (July-August)

Sentiment analysis of tweets directed at US Airlines:

• This task was much more loosely defined, with the main goal being simply to produce a classifier to predict the sentiment of a tweet directed at an airline (positive, negative or neutral).
• To solve this problem, we created a few baseline models, and a final Recurrent Neural Network model.
• This task's other requirement was to get insights from the dataset which could be then used to make recommendations to businesses.
• This involved a significant amount of exploratory analysis in order to gather these insights into the dataset.
• We then produced another presentation detailing our findings.

I worked on a few things in this task:

• Extracting locations from tweet texts using a keyword-based system, and reference lists of airport codes and city names.
• Latent Semantic Analysis (automatic categorisation of data) in order to find the topics that appear most in the dataset.
• A baseline Naive Bayes classifier, implemented using sklearn.