You are here:
About Models
A model is the sophisticated, custom mathematical construct that Einstein Discovery uses to predict a particular outcome. A model accepts inputs (one or more explanatory variables) and produces outputs (a predicted outcome, top factors, and improvements). Create a model and Einstein Discovery generates the predictive and prescriptive analytics.
Note Einstein Discovery stories are now models. We wish we could snap our fingers to update
the name everywhere, but you can expect to see the previous name in a few places until we
replace it.
Models for Predictive and Prescriptive Analytics
Based on data mining, machine learning, and predictive statistical modeling:
- Predictive analytics is the practice of predicting future outcomes based on a comprehensive analysis of past outcomes.
- Prescriptive analytics is the practice of suggesting ways in which to improve your predicted outcomes (improvements).
Einstein Discovery uses the model to generate diagnostic insights, predictions, and improvements.
Terminology
Refer to the following terminology when working with models.
| Term | Definition |
|---|---|
| model | The sophisticated, custom equation that Einstein Discovery generates based on a comprehensive, statistical understanding of past outcomes. Einstein Discovery uses models to predict future outcomes. A model accepts the values of one or more predictor variables as input and produces a predicted outcome as output, along with (optionally) top factors and improvements. |
| predictor | An explanatory variable that a model accepts as input in order to calculate a prediction. Predictors are also known as predictor variables or independent variables. |
| prediction | A derived value, produced by a model, that represents a possible future outcome. You can think of a prediction as the output of a predictive model that is based on the inputs of predictor variables that the model accepts. |
| top predictor | A condition that most significantly drives the predicted outcome. A condition is a data value associated with a variable. In Einstein Discovery, a predictor consists of one or two conditions. |
| improvement | A suggested action that a user can take to improve the likelihood of a desired outcome. Improvements are associated with actionable variables, which are variables over which users can possibly control or influence, such as the shipping method or a subscriber’s membership level. By taking the actions that Einstein suggests, users can increase their chances of having a more favorable outcome. |
| prediction definition | A container object in Einstein Discovery associated with one or more models. If a prediction definition contains multiple models, then each model produces predictions for a different segment of the data. A prediction definition can contain up to ten active models. |
| segmentation | Involves deploying models that target different segments (subsets) of your data. For example, suppose your data contains large, medium, and small customers, and your company organization is oriented around customer size to address the specialized needs of each group. You could build and deploy separate models for large, medium, and small customers to address the unique characteristics of each group. You define segments using filters that specify conditions for each group. Segmentation involves prediction definitions with multiple models. |
| prediction column | In a CRM Analytics dataset, the column where Einstein Discovery stores prediction values returned from the model. |
Types of Models
Einstein Discovery uses thee types of models. The model type depends on the outcome variable used in your story.
| Use Case | Description |
|---|---|
| Numeric Use Case | Numeric fields (measures) can contain many different types of values. Predicting a number field is a regression problem with its own set of metrics to measure model quality. |
| Binary Classification Use Case | Categorical (text) fields (classifications) contain only two qualitative values. Examples include variables that are either true or false, public or private, churned or not churned, and so on. These fields separate your data into two distinct groups. Predicting a categorical field is a binary classification problem with its own set of metrics to measure model quality. |
| Multiclass Classification Use Case | Categorical (text) fields (classifications) with 3-10 possible classes (outcomes). For example, a manufacturer can predict, based on customer attributes, which of six service contracts a customer is most likely to choose. |
Considerations When Working With Models in Einstein Discovery
- For a numeric variable, if Einstein Discovery finds low cardinality (ten or fewer unique values) during analysis, the data type for this variable in the generated model is text rather than numeric.
Did this article solve your issue?
Let us know so we can improve!
