The Predictor, in MindsDB’s words, means the machine learning model. In this dashboard, you will train the models from the Datasources.
How to train Predictor?
MindsDB Scout offers two options for Predictor training:
- Upload the already existing model.
- Train a new one.
To Train New Predictor, the required options are the Predictor Name and the columns to be predicted (target variables).
Inside the Advanced Mode, you can specify additional options before training new Predictor as:
Sample Margin of Error: the amount of random sampling error.
Stop Training After: stop the model training after
Backend: the Machine Learning backend to use for training (Lightwood or Ludwig).
Use GPU: use Graphics Processing Unit for training.
Select columns to be removed for training: speed the training or improve accuracy by ignoring un relevant columns.
The Predictor Results dashboard, provides powerful insights related to the model. The 3 important questions that this section answers are:
- How accurate is the model?
- What is relevant for this model?
- When can you trust this model?