The Role of PMML in Decision Automation
Predictive Model Markup Language (PMML) is an open-source XML-based standard developed by the Data Mining Group (DMG) in order to represent predictive models.
Using this standard, it is possible to construct algorithms in any language (e.g., Python, R, or any other application) and then create a predictive model using PMML.
This is a popular approach for the following reasons:
- Interoperability: PMML allows you to share and reuse your models across multiple applications. There is no need to use a specific language or tool to create a model and then connect it with your applications.
- Available offline: Since it is stored as a single XML file, there is no need for a Web interface to connect the models. Instead, you can use the file offline.
Predictive analytics has leveraged advances in Artificial Intelligence (AI) and Machine Learning (ML) and now plays a major role in modeling and automating decisions. At its core, it facilitates the modeling and definition of decisions that provide real business value to organizations. Essentially, it operationalizes the AI/ML approach to typical decision automation within organizations.
Once the PMML model is designed, it can simply be ‘plugged in’ to existing decision models as part of the end-to-end decision automation process.
So how does it work in practical terms? As a simple example, the following snippet was taken from a UCI ML Breast Cancer Wisconsin (Diagnostic) dataset. It contains a set of data from a breast cancer screening that includes a column that indicates whether or not cancer cells were detected in a particular test:
Based on the dataset shown above, a Gaussian Naive Bayes Classifier was used to train and create a predictive model in Python. This model can be used to predict the possibility of cancer based on patient screening data. Using that predictive model, the developers were able to create a PMML file similar to the snippet shown below:
This PMML file can then be used across multiple applications. A good example of this would be its use in a Decision Requirement Diagram (DRD), as demonstrated in the next section.
Application of PMML as a Part of a Decision Model
The model below showcases the use of a PMML file via the FlexRuleTM Advance Decision Management Suite. This example utilizes inputs from a patient’s breast cancer test and determines whether there is a possibility that cancer exists.
As an output, it shows whether cancer cells were detected, along with the probability of the disease being present.
To see the complete tutorial, visit our Resource Hub.
This output is designed to enable the user to make further decisions, such as whether to contact the subject of the test immediately and/or the type of treatment required, by creating a complete decision model. The complete output can be used as a historical data collection source that enables the construction of a more robust system.
Conclusion
In order to function accurately, predictive models rely on a constant source of recently updated data. Business domains such as healthcare, finance, and education have endless opportunities to improve operational decisions outcome using the latest information and data.
PMML can play a critical role in enabling decision automation scenarios to consume predictive models as part of end-to-end decision automation as well as helping to operationalize AI/ML projects. In this way, PMML can provide real business value to organizations.
To learn more, visit https://www.flexrule.com/archives/pmml-decision-automation/
