Which estimator is best for my machine learning model?

Answer:

While it depends on the machine learning model you are running and how you like to consume your results, the ideal estimator is the one that performs the best without being overfit to the data. If no overrides are made, One AI will go through an intelligent selection of the available estimators (see below) to find the best fit for your model. If you want your model to use a specific estimator or only consider a select few, you can configure this from the augmentation page in the Estimator Configuration section of the One AI configuration (One AI >Edit > One AI Configuration > Estimator Configuration > Override). You can also configure the estimators parameters and hyperparameters here. In the example in the image below, only the LightGBM and SVC estimators would be tried. 

Estimator Configuration

Here are the estimators One AI has available on site:

Classification Estimators:

• AdaBoostClassifier*
• DecisionTreeClassifier
• KNeighborsClassifer
• LightGBM*
• LogisticRegression*
• RandomForestClassifier*
• SVC

Regression Estimators:

• AdaBoostRegression
• DecisionTreeRegressor
• ElasticNet*
• GaussianProcessRegressor
• HuberRegressor
• Lasso*
• LightGBMRegressor*
• LinearRegression*
• RandomForestRegressor*
• Ridge
• SGDRegessor
• SVR

*Indicates that the estimator is tried by default; others should be configured to be tried

For more information about each classifier, search “estimator configuration” in this advanced configuration settings help article.