class monad.ui.interpretability.TreemapHierarchy
monad.ui.interpretability.TreemapHierarchyTreemap parameters to define hierarchy on a plot.
Methods
TreemapHierarchy.__init__(self, interpretability_files_path=None, hierarchy=None)
TreemapHierarchy.__init__(self, interpretability_files_path=None, hierarchy=None)Inits TreemapHierarchy object.
| Example |
|---|
The first few lines of csv file and code snippet bellow demonstrate how to generate Treemap the with predefined hierarchy.
department_name, section_name, colour_group_name, article_id, prod_name
Jersey Basic, Womens Everyday Basics, Black, 0108775015, Strap top
Jersey Basic, Womens Everyday Basics, White, 0108775044, Strap top (1)
Jersey Basic, Womens Everyday Basics, Off White, 0108775051, Strap top (2)
Clean Lingerie, Womens Lingerie, Black, 0110065001, OP T-shirt (Idro)
Clean Lingerie, Womens Lingerie, White, 0110065002, OP T-shirt (Idro)from pathlib import Path
from monad.ui.interpretability import TreemapGenerator, TreemapHierarchy
hierarchy = TreemapHierarchy(
hierarchy_path=Path("<path/to/csv/file/with/hierarchy.csv>"),
levels=["department_name", "section_name", "colour_group_name", "article_id"],
feature_values_importance_path=Path("<path/where/interpret/results/were/saved>", "transactions", "article_id", "values_importance.json"),
entity_name_column="prod_name",
)
tg_from_hierarchy = TreemapGenerator(hierarchy=hierarchy)
tg_from_hierarchy.plot_treemap(
output_file_path=Path("churn_hierarchy_treemap.html"),
max_depth=4,
)Te code above will result in the chart below with the hierarchy of levels as follows: department_name -> section_name -> colour_group_name -> article_id / prod_name
| Parameters |
|---|
hierarchy_path: pathlib.Path
Path to a csv file that defines hierarchy. It should contain columns corresponding to the levels of hierarchy on a plot. Each row should define mapping from the last hierarchy level to other levels. The column with the last hierarchy level should contain unique values.
Example content of a file is demonstrated by a table below which corresponds to the hierarchy ["level_1_category", "level_2_brand", "level_3_product_id"].
| level_1_category | level_2_brand | level_3_product_id | product_name |
|---|---|---|---|
| category A | brand A | 1 | product 1 |
| category A | brand B | 2 | product 2 |
| category A | brand B | 3 | product 3 |
| category B | brand A | 4 | product 4 |
levels: List[str]
List of hierarchy levels that should be used on a treemap. From the most general level to the most specific one. Usually last hierarchy level is product id. Last hierarchy level should be in the model training data, other levels are independent of the training data.
feature_values_importance_path: pathlib.Path
Path to a JSON file with feature value importances corresponding to feature values from the last hierarchy level. File with importances should be one of the files generated with the interpret function described in Interpreting your model's predictions.
entity_name_column: Optional[str]
Default: None
Column name from csv file with hierarchy path. This column should contain names corresponding to the last hierarchy level that should be used on the Treemap. If None, original values of the last hierarchy level will be used.