House Price Prediction with Neighbourhood Graph Features
A group coursework project predicting California housing prices using feedforward neural networks. Enhanced the baseline model by engineering neighbourhood features from spatial graph relationships (k nearest neighbours), achieving a 35% reduction in test error.
A natural language powered OS automation agent. Uses OCR, SQLite memory, action planning and Python scripting to execute tasks on macOS locally without browser automation.
Analysis of how hyperparameter choices affect evolutionary optimisation performance by running controlled trials across population sizes and mutation rates, evaluating convergence behaviour through statistical metrics and visualisations.
A Python search engine emulator that crawls websites, stores link graphs in SQLite, computes PageRank iteratively and includes a D3.js force directed visualiser for further exploration.
An evolutionary algorithm that transforms random text strings into a target Shakespearean sonnet using mutation, crossover, scoring functions and iterative optimisation.