Next generation analytics
Machine Learning that is cost-effective, practical and fast to implement.
Our ML solution is the result of our analytics team’s dedicated investigation of machine learning technologies that are practical, cost effective and compliant to modelling standards expected by business. It is the combination of technologies in both analytics and data management that brings our solution to life, referred to as the “Machine Learning Platform” (MLP).
How it Works
Our solution can ingest data from multiple sources, wrangle data for modelling needs and store modelling outcomes in the data management platform. Execution of the data models rests with Stratus – Principa’s propriety execution solution that incorporates model management and audit features. Both these are critical features in a business ready machine learning solution. With this functionality in place, models can be retrained quickly and cost effectively, based on client needs and application use. Standard or customisable reports continually monitor modelling outcomes.
The solution can be deployed on client premise or a virtual server in the cloud. Our ML offering is highly customisable based on client needs and may be deployed in part or as a complete end to end solution.
Quick Step ML Model Building
We developed our own flexible approach to ML model building in terms of which we build what is referred to as “Quick Step” models. Our approach allows us to develop multiple concurrent models on one set of data.
During a recent model building exercise, six different modelling techniques were used with three or four different parameter settings. This resulted in twenty-two different scores. Each score was assessed for its discriminatory strength and for over-fitting, to determine the most suitable model for the specific business problem. Currently, Quick Step models are trusted with over 70-billion of our scores, calculated annually.
Machine Learning Platform (MLP)
We can assist in deploying your machine learning models on our proprietary Machine Learning Platform. The solution is deployable on-premise or in the cloud. Our MLP covers the ETL process through our data management solution. We use various technologies to achieve this, dependent mostly on whether it’s an on-premise or cloud deployment.
ML model management
Once ML models are built they need to be deployed and managed. This refers to the monitoring of the stability of the data and model performance. For this you need an experienced team that understands the various models, the data as well as the metrics that might indicate redevelopment. If a ML environment is built correctly, data should be modelling ready and rapid redevelopment is possible.
Our data science team has extensive experience in managing ML environments and models.