Basic predictive analysis glossary
Predictive analysis – a class of data analysis methods that enables the prediction of object future behavior under given conditions. It uses statistical methods to analyze current and historical data and make a prediction about future events.
Lookup value prediction – one of the tools for predictive data analysis, which enables you to predict the value of the lookup field based on the analysis of existing data in the system. Read more >>>
Predictive scoring – grades Creatio records based on historical and current data. Read more >>>
Numeric value prediction – enables you to predict the value of a numeric field based on the analysis of existing Creatio data. Read more >>>
Machine learning problem – a set of instructions, which describes a problem that must be solved by predictive analysis. The list of problems is available in the [ML problem types] lookup.
Machine learning model – an algorithm that defines the data, which is a basis for the predictive analysis of solving a machine learning problem. The list of machine learning models is available in the [ML models] section.
Machine learning model instance – a set of patterns obtained by the machine learning model as a result of processing the historical data.
Historical data – a collection of data obtained from system records to create a model. The historical model takes into account records that were created and populated with data before starting a new machine learning model instance.
Model training – a process, during which the machine learning model processes historical data to identify patterns that enable it to solve a particular machine learning problem. A new machine learning model instance is created during model training. Use the [ML models] section to specify the model retraining frequency.
Prediction – a list of possible values of a lookup field including their probability. The probability of prediction is indicated in percentage (rounded up to integers) for each lookup value.
Quality metric lower limit – the prediction probability threshold that has to be met to use the prediction service. The instances that do not meet the quality metric lower limit are not used by Creatio and are placed into a queue for retraining. We do not recommend setting up the quality metric lower limit to a value lower than 0.50. You can modify the model quality metric lower limit value in the [ML models] section.
See Also
•Numeric field value prediction
