You are here:
Supported Predictive Model Types
Predictive models address common business use cases.
- Regression for Numbers
Numeric (regression) use cases target numeric outcomes represented as quantitative data (measures), such as currency, counts, or percentages. - Binary Classification for Text
Binary classification solutions target business outcomes with only two possible results, which are represented as text data. - Multiclass Classification
Multiclass classification solutions provide detailed predictions, closer alignment with real-world situations, and the ability to take targeted actions.
Regression for Numbers
Numeric (regression) use cases target numeric outcomes represented as quantitative data (measures), such as currency, counts, or percentages.
Example of Regressions
Here are just a few examples of how models created in AI Models (formerly Einstein Studio) can help you improve numeric outcomes in your organization:
- Predicted Amount of an opportunity
- Predicted Time-to-Close of an opportunity
- Predicted Customer Lifetime Value of an account
- Predicted Customer Satisfaction of a case
Supported Algorithms for Regression
- Generalized Linear Model (GLM) - Linear Regression
- XGBoost
- Gradient Boost Machine (GBM)
Binary Classification for Text
Binary classification solutions target business outcomes with only two possible results, which are represented as text data.
Binary Outcomes
These outcomes are typically yes/no questions that are expressed in business terms, such as churned or not churned, opportunity won or lost, employee retained or not retained, and so on. For analysis purposes, AI Models (formerly Einstein Studio) converts the two values into boolean true and false.
Example Binary Classification Solutions
Here are just a few examples of how models created in AI Models can help you improve binary classification outcomes in your organization:
- Predict the probability to win an Opportunity
- Predict the probability for an Account to buy a specific product
- Predict the probability that a Lead will convert
- Predict the probability that an Account will churn
Supported Algorithms for Binary Classification
- Generalized Linear Model (GLM) - Logistic Regression
- XGBoost
- Gradient Boost Machine (GBM)
Multiclass Classification
Multiclass classification solutions provide detailed predictions, closer alignment with real-world situations, and the ability to take targeted actions.
Multiclass Outcomes
With AI Models (formerly Einstein Studio), you can use multiclass classification to predict the likelihood that records fall into one of 3 and up to 50 buckets. In contrast with binary classification, which provides two outcomes, yes or no, multiclass models can handle multiple possible outcomes.
Example Multiclass Classification Solutions
Improve outcomes in your organization with these multiclass use cases.
- Predict the most likely product from multiple up-sell or cross-sell options
- Predict the most likely product category from a catalog of offerings
- Predict the most likely next sales stage in the opportunity lifecycle
- Classify support cases into one of several case reason categories
- Classify accounts into one of several predefined customer segments
For example, a marketing team uses multiclass classification to predict the most effective communication channel for a specific customer. To target the campaign for a specific product, the model evaluates the probability that a record belongs to various categories. When customers use touchpoints, such as phone, email, or chat, the model analyzes their demographics and past interactions to recommend the appropriate channel. Precise targeting ensures that the optimal marketing channel reaches customers to drive a purchase, an upsell, or increased brand loyalty.
