Topic elementary workflow of data processing for common tutorial datasets:
- processing of the iris datasets
- one-hot encoding
- processing of the titanic data
- data preprocessing
- one-hot representation: cardinal vs. ordinal veriables
- processing of the fraud data
- imbalanced class probelm
- potential solution using data upsampling
Data for tasks 1 and 2 are readily available. Fraud data are provided in the /data folder.
Hands on analysis of the real-world data provided by colleagues from RTI, see folder /data. The task is to predict column "Ra" using columns "lpow;lspeed;dist;vol".
Topics to discuss:
- feature engineering / feature selection
- residue analysis - how to understand the results
- model selection