README.md

---
title: ISCO-08 Hierarchical Accuracy Measure
datasets:
-  ICILS/multilingual_parental_occupations
tags:
- evaluate
- metric
description: "The ISCO-08 Hierarchical Accuracy Measure is an implementation of the measure described in [Functional Annotation of Genes Using Hierarchical Text Categorization](https://www.researchgate.net/publication/44046343_Functional_Annotation_of_Genes_Using_Hierarchical_Text_Categorization) (Kiritchenko, Svetlana and Famili, Fazel. 2005) applied with the ISCO-08 classification scheme by the International Labour Organization."
sdk: gradio
sdk_version: 3.19.1
app_file: app.py
pinned: false
---

# Metric Card for ISCO-08 Hierarchical Accuracy Measure

## Metric Description

The measure accounts for the hierarchical structure of the ISCO-08 classification scheme by rewarding partial correctness in classification, penalizing distant errors, and applying higher penalties for higher-level errors. It is designed to evaluate the accuracy of predictions within a hierarchical occupation classification system, such as the ISCO-08 scheme, which organizes occupations into a tree structure with four levels: major group, sub-major group, minor group, and unit group.

1. Partial correctness in classification: Misclassification within the same major group (e.g., within "Professionals") is penalized less than misclassification across major groups (e.g., from "Professionals" to "Managers"). This acknowledges the similarity in skill level and specialization within major groups.
2. Penalization of distant errors:
 2.1 Correct classification into a more specific group (e.g., "Medical doctors" within "Health professionals") is preferred over a broader classification (e.g., the entire "Health professionals" group).
 2.2 Conversely, incorrect classification into a more specific but unrelated group (e.g., "Nursing and midwifery professionals" when the correct classification is "Engineering professionals") is penalized more heavily than an incorrect but broader classification.
3. Higher penalties for higher-level errors: Misclassifications at the major group level are more severely penalized than errors within sub-major, minor, or unit groups, reflecting the greater disparity in skill level and specialization.
4. Extension with ancestors in the hierarchy: For the hierarchical precision ($hP$) and recall ($hR$) calculations, an occupation's classification is extended to include not only its specific group but also the parent groups to which it belongs, excluding the root level. This allows for an evaluation that recognizes partial correctness based on the occupation's position within the ISCO-08 hierarchy.

### Original description

1. The measure gives credit to partially correct classification, e.g. misclassification into node $I$ (ISCO unit group "1120") when the correct category is $G$ (ISCO unit group "1111") should be penalized less than misclassification into node $D$ (e.g., ISCO unit group "1211") since $I$ is in the same subgraph (ISCO sub-major group "11") as $G$ and $D$ is not.
2. The measure punishes distant errors more heavily:
    1. the measure gives higher evaluation for correctly classifying one level down compared to staying at the parent node, e.g. classification into node $E$ (ISCO minor group "111") is better than classification into its parent $C$ (ISCO sub-major group "11") since $E$ is closer to the correct category $G$; and
    2. the measure gives lower evaluation for incorrectly classifying one level down comparing to staying at the parent node, e.g. classification into node $F$ (ISCO unit group "1120") is worse than classification into its parent $C$ since $F$ is farther away from $G$.

The features described are accomplished by pairing hierarchical variants of precision ($hP$) and recall ($hR$) to form a hierarchical F1 ($hF_β$) score where each sample belongs not only to its class (e.g., a unit group level code), but also to all ancestors of the class in a hierarchical graph (i.e., the minor, sub-major, and major group level codes).

Hierarchical precision can be computed with: `$hP=\\frac{| \\v{C}_i ∩ \\v{C}^′_i|} {|\\v{C}^′_i|}=\\frac{1}{2}$`

Hierarchical recall can be computed with:
`$hR = \frac{| \v{C}_i ∩ \v{C}^′_i|}  {|\v{C}_i |} = \frac{1}{2}$`

Combining the two values $hP$ and $hR$ into one hF-measure:
`$hF_β = \frac{(β^2 + 1) · hP · hR}{(β^2 · hP + hR)}, β ∈ [0, +∞)$`

## How to Use
*Give general statement of how to use the metric*

*Provide simplest possible example for using the metric*

### Inputs
*List all input arguments in the format below*
- **input_field** *(type): Definition of input, with explanation if necessary. State any default value(s).*

### Output Values

*Explain what this metric outputs and provide an example of what the metric output looks like. Modules should return a dictionary with one or multiple key-value pairs, e.g. {"bleu" : 6.02}*

*State the range of possible values that the metric's output can take, as well as what in that range is considered good. For example: "This metric can take on any value between 0 and 100, inclusive. Higher scores are better."*

#### Values from Popular Papers
*Give examples, preferrably with links to leaderboards or publications, to papers that have reported this metric, along with the values they have reported.*

### Examples
*Give code examples of the metric being used. Try to include examples that clear up any potential ambiguity left from the metric description above. If possible, provide a range of examples that show both typical and atypical results, as well as examples where a variety of input parameters are passed.*

## Limitations and Bias
*Note any known limitations or biases that the metric has, with links and references if possible.*

## Citation
*Cite the source where this metric was introduced.*

## Further References
*Add any useful further references.*