Datasets

Data for each day

The dataset for each teaching day. Day 1 uses the official UCI bank marketing file, 41,188 recorded calls with 20 inputs and the target. Columns are separated by semicolons, not commas. Keep the file in a data folder beside the notebook. The remaining days are released before each session.

Source. UCI Machine Learning Repository, Bank Marketing, DOI 10.24432/C5K306, licence CC BY 4.0. Moro, S., Cortez, P. and Rita, P. (2014). The file holds calls made between May 2008 and November 2010.

The 20 input columns fall into four groups. They describe the client, the current contact, the campaign history, and the economic conditions at the time. Two columns need care. duration records how long the call lasted, which is only known once the call has ended, so a model that predicts beforehand cannot use it. In pdays, the value 999 means the client was never contacted before, not that 999 days passed.

D1 Foundations: data, prediction, and trust MondayDay 1
D2 Regression and classification: from models to decisions TuesdayDay 2 To be released
D3 Flexible models and honest evaluation WednesdayDay 3 To be released
D4 Unsupervised learning, modern AI, and responsible use ThursdayDay 4 To be released
Oxford · United Kingdom Teaching CV
University of Oxford