Representative Datasets

Perera Lago, J., Toscano Durán, V., Paluzo-Hidalgo, E., Narteni, S. & Rucco, M. Application of the representative measure approach to assess the reliability of decision trees in dealing with unseen vehicle collision data

Two experiments were developed:

1. Synthetic Dataset,
2. Vehicle Collision Dataset.

In both four sets were considered, the train set, two random subsets generated from the train set and the test test.

In addition, the file correlationEpsilon_Similarity, we repeated the experiment for 100 subsets of Collision Dataset and the Spearman’s correlation (Sp) between the ε-representativeness and the metric of the ordering of the feature importance was computed obtaining significant correlation.

List of main needed libraries

• numpy
• pandas
• scikit-learn

To install all the necessary python(we use Python 3.11.6) dependencies to run these experiments use the next command:

pip install -r requeriments.txt 

Experiments were run on a computer with AMD Ryzen 7 7735HS with Radeon Graphics CPU and a NVIDIA GeForce RTX 4060 with 8 GB of RAM GPU