Modeling educational data (Part 2)

This article is Part 2 of our 2-part series in modeling educational data. In the first part, we introduced network structures and gave an overview of educational mapping.

In this second part of our introductory series, we walk you through how to create a network map of your own educational data. If you want to build a model for your own educational data, read on for how to get started.

Creating your educational network model

First you need to define the entities in your model. Think about the “things” (nouns) that represent your system. Different types of entities can exist in different models. For example, in a curriculum network model, we model as entities the Institution, the Departments within the institution and the Courses within departments.

Entities are represented as vertices in the network model. A vertex is assigned text and numeric attributes that represent information on the entity. For example, the name of the entity can be an attribute.

curriculum map model

Next you need to define the relationships in your model. These relationships define how one entity relates to another entity, so relationships are verbs. Relationships can express organizational relationships (e.g., has-parent-of) or they can express educational relationships (e.g., Course X addresses Outcome A; Outcome B leads-to Outcome C).

Relationships are represented as edges in the network model. An edge between two vertices is a model of the relationship between those two entities. An edge is assigned text and numeric attributes that describe the relationship. For example, we might choose to assign an edge attribute that represents the strength of the relationship on a scale from 0 – 1.

What are some examples of entities in a network model?

Entities could include: CourseDepartmentInstitutionAssessmentSkillTopicOutcomeMicro-outcomeJobResource

Here are some of the entities our educational network models have used:

What are some examples of relationships in a network model?

Relationships could include: has-parent-ofhas-prerequisite-ofaddressesrequiresleads-tois-related-to

Here are some of the entities our educational network models have used:

Granularity of mapping your educational data

A common question is: How do I choose what to model as entities? Should I model topics or outcomes or something else?

What is best depends on what you want to achieve, so start by defining your potential use cases.

For example, if your goal is to use your network model as a foundation for adaptive learning, you will need to define at a granular level “what is a student expected to know or to be able to do.” That means you will need Outcomes and/or Micro-outcomes as model entities.

We have found that linking topics to topics is helpful to students (Crosslinks is based on linking topics to topics and resources to topics) and is a tempting place to start, but it can be ambiguous — for example, what does it mean to link “matrix” to “eigenvalue”? Our more recent mapping work instead links outcomes, where we can more directly capture intent. For example, with an outcome we can model what is it about an eigenvalue that a student needs to be able to do or explain, and how does this measurable skill relate to another skill involving a matrix?

Building an educational network model is an iterative process. As your map starts to take shape, you may realize that you need to introduce a new class of entity or relationship. Also keep in mind that attributes can be assigned to both entities and relationships, but be careful not to use attributes when introducing a new class of entity leads to a more scalable model. For example, we could define a Course as an entity and assign it an attribute called department. That works fine for tagging each Course with its own department. But now what if a course is cross-listed under two departments? The attribute gets tricky to manage. A more scalable way to model this setup is to define Department as an entity class and introduce the relationship has-parent-of. Then every Course can be tagged with its parent Department(s). These relationships are now represented explicitly in the network model, making them adaptable and easy to analyze.

Where to go from here? You can walk through our a mapping example where we take you from PDF to map. For more a more detailed write-up, read our journal paper on network models for education.