Starting with the data
Real-world datasets rarely arrive ready for modelling. This work focused on understanding the structure and limitations of a dataset before selecting analytical techniques, including the identification of missing values, inconsistent formats, unusual observations and variables requiring transformation.
Preprocessing & exploration
Data preparation was treated as an analytical decision rather than a mechanical step. Exploratory statistics and visualisation were used to identify distributions, relationships and potential quality issues before model development.
- Assessed data types, completeness and consistency.
- Investigated distributions and anomalous values.
- Prepared variables for subsequent analysis.
- Used exploratory analysis to guide method selection.
Machine learning & evaluation
Machine-learning techniques were applied within the wider KDD workflow and evaluated using more than predictive performance alone. Suitability, assumptions, interpretability and the quality of the underlying data were also considered.
Transferable capability
The project demonstrates analytical thinking, data quality awareness, evidence-based method selection, Python development and the ability to communicate why an analytical approach is appropriate rather than simply applying a model without context.