Samples count report

An important dataset property to keep an eye on when aiming for modeling is the number of samples for each class available in the datasets.

You can easily recover this using the datasets samples-count subcommand.

For instance, the following command computes the samples count for the unfitered version of the ucdavis-icdm19 dataset.

tcbench datasets samples-count --name ucdavis-icdm19

Output

unfiltered
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━┓
┃ partition                   ┃ app           ┃ samples ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━━━┩
│ pretraining                 │ google-doc    │    1221 │
│                             │ google-drive  │    1634 │
│                             │ google-music  │     592 │
│                             │ google-search │    1915 │
│                             │ youtube       │    1077 │
│                             │ __total__     │    6439 │
├─────────────────────────────┼───────────────┼─────────┤
│ retraining-human-triggered  │ google-doc    │      15 │
│                             │ google-drive  │      18 │
│                             │ google-music  │      15 │
│                             │ google-search │      15 │
│                             │ youtube       │      20 │
│                             │ __total__     │      83 │
├─────────────────────────────┼───────────────┼─────────┤
│ retraining-script-triggered │ google-doc    │      30 │
│                             │ google-drive  │      30 │
│                             │ google-music  │      30 │
│                             │ google-search │      30 │
│                             │ youtube       │      30 │
│                             │ __total__     │     150 │
└─────────────────────────────┴───────────────┴─────────┘

While to obtain the breakdown of the first train split

tcbench datasets samples-count --name ucdavis-icdm19 --split 0

Output

filtered, split: 0
┏━━━━━━━━━━━━━━━┳━━━━━━━━━┓
┃ app           ┃ samples ┃
┡━━━━━━━━━━━━━━━╇━━━━━━━━━┩
│ google-doc    │     100 │
│ google-drive  │     100 │
│ google-music  │     100 │
│ google-search │     100 │
│ youtube       │     100 │
├───────────────┼─────────┤
│ __total__     │     500 │
└───────────────┴─────────┘

...or the human test split

tcbench datasets samples-count --name ucdavis-icdm19 --split human

Output

filtered, split: human
┏━━━━━━━━━━━━━━━┳━━━━━━━━━┓
┃ app           ┃ samples ┃
┡━━━━━━━━━━━━━━━╇━━━━━━━━━┩
│ youtube       │      20 │
│ google-drive  │      18 │
│ google-doc    │      15 │
│ google-music  │      15 │
│ google-search │      15 │
├───────────────┼─────────┤
│ __total__     │      83 │
└───────────────┴─────────┘