Dataset

The dataset selected is the Electrical Grid Stability Simulated Data.

Features after preprocessing

• Binary classification (last column): 1 = stable | -1 = unstable
• 10,000 data points (3,620 stable and 6,380 unstable)
• 12 real inputs
• Shuffled
• 20% for testing and 80% for training

Important notes

• Since this dataset is imbalanced, we will use the F1-score instead of the binary loss.
• The last column of the dataset, i.e., the class label, was replaced for 1 = stable | -1 = unstable
• The last column of the inputs, i.e., index 13, was removed since it is equal to the class itself and has no use in this analysis.

Structure

dataset
│   └──README.md
│   └──split_dataset.m
│
└───training
│   └──dataset_training.txt
│   
└───testing
│   └──dataset_testing.txt  
│ 
└───charts
    └──bar_class.eps
    └──bar_class.png