Predictive Analytics

To calculate predictive analytics online, simply copy your data into the table above and select the variables for your model. First select the method, then choose one dependent variable and the independent variables that should be used as predictors.

If the dependent variable is categorical, numiqo calculates a classification model. If the dependent variable is metric, numiqo calculates a regression model. This allows you to predict categories such as customer groups or defect types, or metric values such as sales, costs, or defect rates.

The available methods are the Decision Tree calculator, Random Forest calculator, Gradient Boosting calculator, and CHAID as an additional decision tree method for categorical dependent variables.

Calculate a predictive model

For example, if you want to predict whether a product will be defective, select the defect status as the dependent variable and select the production variables as predictors. numiqo then creates a classification model and shows the results clearly presented.

If you want to predict a metric value, for example a defect rate or a sales value, select a metric dependent variable. The calculator then creates a regression model and reports prediction errors such as RMSE and MAE where available.

After a CART Decision Tree, Random Forest, or Gradient Boosting model has been calculated, you can send it to the model prediction page. There you enter values for the selected predictors and receive the predicted category or predicted numeric value.

Predictive analytics methods

Decision trees are useful when you want an interpretable model with simple rules. Random Forest combines many trees and is often more stable than a single decision tree. Gradient Boosting builds trees step by step and can perform well when the settings fit the data. CHAID is useful for categorical dependent variables and segmentation tasks.

If you are unsure which method fits best, use Compare Models to evaluate CART Decision Tree, Random Forest, and Gradient Boosting on the same validation split. This helps you compare accuracy or error measures before choosing a model for prediction.

For an introduction to the method, see the Predictive Analytics tutorial. It explains classification, regression, influencing factors, and first prediction models in numiqo.