> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.umnai.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.umnai.com/_mcp/server.

# Identifying data segments

Umnai can identify data segments by inspecting the rules created during induction.

During [induction](../induction), the training dataset is segmented into regions that behave similarly. These [segments](../explanations/global-explanations/dataset#segments) are encoded in the `conditional_expression` of each [rule](../explainable-neural-net-xnn#rule).

This information helps explain both the dataset and the model: it shows how Umnai grouped data by behaviour, and how those groups relate to the segments people already use in the domain.

## How segments are represented

Rules describe the conditions that activate a partition.

The `conditional_expression` of a rule defines the segment of data associated with that rule. For example, an `Age` rule may define a range such as:

> `0 < Age <= 23`

When a query falls within that range, the corresponding rule and partition can activate.

Looking across the rules in a module shows how induction segmented that feature or interaction into regions of similar behaviour.

## Comparing discovered segments with business segments

Discovered segments can be compared with existing business, operational, or domain-defined segments.

For example, a marketing team may use standard `Age` groups:

| Business segment          | Range      |
| ------------------------- | ---------- |
| Children and young adults | `0 to 18`  |
| Adults                    | `18 to 65` |
| Older adults              | `over 65`  |

The `Age` module rules may identify different behavioural bands:

| Discovered segment | Range      |
| ------------------ | ---------- |
| Segment 1          | `0 to 23`  |
| Segment 2          | `23 to 55` |
| Segment 3          | `over 55`  |

The difference between these segmentations can be useful. It may show that the model has learned behavioural boundaries that do not match the organisation's existing segmentation.

## Using segment insights

Segment insights can support business and model review.

If the discovered segments make sense, they may strengthen confidence in the model and in the dataset. If they differ from expected business segments, they may prompt further investigation.

For example, the marketing team may re-evaluate its market segmentation and campaigns if the model consistently identifies `0 to 23`, `23 to 55`, and `over 55` as behaviourally meaningful groups rather than the existing `0 to 18`, `18 to 65`, and `over 65` groups.

This does not automatically mean the original business segmentation is wrong. It provides evidence that can be compared with domain knowledge, business strategy, and additional data.