Skip to main content
All CollectionsFAQModeling FAQGeneral Modeling FAQ
What types of models can I build with the G2M platform?
What types of models can I build with the G2M platform?

This article discusses what types of machine learning models you can build with the the G2M platform.

Updated over 5 months ago

The G2M platform allows you to build models that fall into these broad categories:

  • Propensity models: also known as classifiers, propensity models are a type of supervised machine learning and usually predict a binary outcome such as 'Will this prospect buy my product?' or 'Will this customer cancel her service?' The algorithms available in the G2M platform to build a propensity model include:

    • Logistic regression

    • LightGBM

    • Random forest

    • Gradient boosting

    • XGBoost

    • Adaptive boosting

    • Extra trees

  • Clustering models: clustering models are a type of unsupervised machine learning and usually identify natural groupings, segments, or clusters in your data. Analysts use clustering to group prospects and customers into similar groups for the purpose of customer segmentation. It can also be applied to other use cases such as segmenting survey responses, operational events, etc. The algorithms available in the G2M platform to build a clustering model include:

    • K-Means

    • PCA / K-Means

    • BIRCH

    • DBSCAN

    • Gaussian mixture

    • Hierarchical agglomerative

    • Mean shift

    • OPTICS

    • Spectral clustering

  • Regression models: regression models are a type of supervised machine learning and usually predict a numerical variable such as total sales using a broad variety of inputs. The algorithms available in the G2M platform to build regression model include:

    • Linear regression

    • Random forest regression

    • Gradient boosting regression

    • XGBoost regression

    • LASSO regression

    • Ridge regression

    • Bayesian ridge regression

  • A/B testing models: A/B testing models are a type of analysis that compares outcomes by adjusting for a number of controlled variables so the comparison is as apples-to-apples as can be:

    • Propensity score matching

    • Propensity score blocking

    • Propensity score stratification

    • Propensity score weighting

    • Ordinary least squares

  • Marketing mix models: marketing mix models are a type of specialized regression that accounts for lagging and saturation effects commonly found in marketing mix modeling problems:

    • Bayesian regression with adstock

    • Bayesian regression with Hill adstock

    • Bayesian regression with carryover


Did this answer your question?