The consensus model is built on the top of individual models.
The individual models should be developed using the same dataset and same pre-processing options. These models should be saved before they are used as part of the consensus (open each model that you would like to click respective button to save it). The saved models are visible inside of "Models/Apply a model" browser.
After that start to create a model using "Model/Create a model" menu and follow the respective steps.
On the final step you can select averaging options.
- Simple average
- Optimal combination of models for each property
- Weighted by model accuracy
- Weighted by molecule prediction accuracy
- Best model for molecule
The most advanced is "Optimal combination of models for each property", which finds best combination of models for property (each property in case of multi model task)
N.B.! Use exactly the same pre-processing options as those used for the individual model since, otherwise, model cannot be created. If you are not sure, before creation of the consensus model, select create a model using "use another model as a template" at the bottom of the page
"You can create a model from template: import an XML model template or use another model as a template"
and use one of individual model as the template. Once template is selected, deselect
Skip model configuration and use the predefined settings (otherwise the template model will be just duplicated), select consensus model and continue with its development.
In this way model will set all pre-processing options exactly as they were used for the individual model.