Overview
In this section, a random forest classifier is applied to your embedded data, generating your VPS results. The configurations in this section involve telling the system how you want your data to be clustered for learning/testing purposes, which may vary depending on your datasets. In other words, here you are controlling the training and test groups that are used for validation of the model.
Read on for more context and explanations about what this step entails.
Key Concepts by Parameter
These settings help the model capture the right amount of detail, preventing it from being too specific (overfitting) or too general (underfitting) and improving its performance on new data.
Parameter: Minimum Cluster Size
The Minimum Cluster Size determines how far the clustering algorithm will search to group data points into clusters.
This parameter influences the scale of the clustering process, whether the algorithm focuses on nearby points to capture finer details or considers points farther apart for a more generalized grouping.
Adjusting this setting helps the model align with the complexity of the geologic setting.
Parameter: Minimum Samples
The minimum samples parameter specifies the smallest number of data points needed to form a cluster.
This setting influences the model's sensitivity to data density.
A lower number allows for detecting smaller clusters in varied settings, while a higher number filters out smaller clusters to focus on larger patterns in more uniform settings.
Adjusting this setting helps the model align with the complexity of the geologic setting and/or capture the subtle differences in the mineral systems between different mineralized zones.
Tips:
If your geologic setting is highly varied, use smaller groups with a lower minimum distance and minimum samples to capture details in order to create more clusters.
If your geologic setting has little variation, use larger groups with a higher minimum distance and minimum samples to capture broader patterns.
Parameter: Advanced Features
To learn more about Advanced Features for Building Predictive Models, click here.
Still have questions?
Reach out to your dedicated VRIFY AI contact or email Support@VRIFY.com for more information.