The purpose of predictive analytics is to forecast future events. Decisions based on analytics are more likely to be correct than decisions based on intuition. The field of Predictive Analytics encompasses a variety of statistical techniques from modeling, machine learning and data mining that analyze current and historical facts to make predictions about future or otherwise unknown events.
Most predictive models define the relationship between a number of input variables and a single output variable. The output is referred to as the dependent variable or target variable. The inputs are referred to as the independent variables or explanatory variables. Data exploration and preparation is often required on both the target and explanatory variables prior to the development of a predictive model.
Model performance indicators
Model quality is measured via model performance indicators.
The two main ones are:
- Predictive Power (Ki): Measures the capacity of the input parameters (the explanatory variables) to explain the target (i.e. the proportion of the target's variability) is a measure specific to SAP but is related to standard model quality indicators in general use. It varies from indicating that a model is a pure random model (Ki tends to 0) or a perfect ideal model (Ki tends to 1)
- Prediction Confidence (Kr): Measures the ability of a model to display the same level of performance on new datasets as it did on the training dataset. It is a measure of the robustness of a model or how generalized its application is. Before a model is used for prediction, its Kr should be >= 0,95.
Both performance indicators have values ranging from 0 to 1.
Key process steps
- Train a model
- Retrain a model if required
- Set a model version to active
- Apply a model to a dataset
Each application has a specific predictive scenario. Use the Predictive Models app to create a model from the existing template based on the predictive scenario. Choose the corresponding predictive scenario and perform the required test steps to train the model.
The predictive model calculates the predictive power and the prediction confidence and evaluates the quality of your model using a quality range from 1 to 5. Based on the quality indicator for your model, you can decide if you want to use your model version. If the model quality needs to be improved, retrain the model. When the training process is complete, the status of the model version will be set to READY.
Decide which model version will be used to generate predictions when the modeling context is queried. Activate this model version.
Apply the model to a data set from your business data, for example, use the Quotation Conversion Rates app to compare the actual and predicted conversion rates of quotations to sales orders.
Quotation Conversion Rates model
Quotation Conversion Rate measures the percentage of the net value of order items that has been converted from a quotation item, based on the total net value of quotation items.
The KPI is calculated at the item level of a quotation. The KPI only considers quotation items that meet the following criteria:
- Net value >0
- Valid on the current date
- Not fully referenced (header of the quotation)
- Overall status is not completed (header of the quotation)
This model scenario provides reliable predictions for the sales manager to monitor the probability of a sales quotation being converted to a sales order, assisting the sales manager to plan more reliably.
- Decreases manual effort for sales planning and delivers more accurate sales forecasts
- Provides reliable predictions for achievable sales volumes
- Increases customer satisfaction by achieving a better delivery performance
- Increases working efficiency by providing better insight to sales managers for decision making
Main used apps
|Process step||Role Name||Fiori App Name|
|Train a model||Analytics Specialist||Predictive Models|
|Set a model version to active||Analytics Specialist||Predictive Models|
|Apply a model to a dataset||Sales Manager||Quotation Conversion Rates- Predicted|