Traditionally advanced analytics was utilized primarily to gain an historical understanding of business performance. Data scientists spent countless hours mining, munging, and model building; wasting expensive resources and valuable time only to get a rear view understanding of business performance.
Starting Now: Use Models to Automate Real-time Decision-making
Until recently it was just not possible to operationalize Machine Learning (ML) models. VoltDB was one of the first database companies to introduce the ability to embed ML models in-database for real-time decisioning. And from today onwards embedded ML just got even easier; users can seamlessly create VoltDB User Defined Functions (UDFs) from their PMML models and use them in SQL and stored procedures for real-time scoring on incoming / streaming or existing data. As necessary; data scientists can train and update the ML models based on new and historical data, and then reload them into VoltDB without sacrificing speed or availability. Frequent model updates ensure that the models/rules are always up to date and relevant. The updated ML model is made available immediately and consistently across the entire VoltDB cluster. With today’s release, VoltDB is the only company that offers the ability to take intelligent decisions on fast streaming data at high throughput and very low latency (milliseconds) by leveraging complex ML models in-database.
VoltDB understands the importance of immediacy, and our engineering team has been working hard to make the product faster, easier to deploy and use. We are very proud to introduce to you VoltML; which allows you to deploy complex Predictive Model Markup Language (PMML) models in production easily and quickly without having to write any code. With VoltML you can use simple SQL queries to take real-time intelligent decisions on all your data. PMML models can be deployed seamlessly in just minutes, allowing for reduced analytical model processing times, frequent model updates, ensuring optimal accuracy and reliable decisions.
The real-time (milliseconds) aspect of ML is becoming increasingly important to a wider range of use case where decisions need to taken instantly by leveraging complex models which query hundreds to thousands of rules. A host of use cases such as: credit card fraud detection, real-time bidding on online ads, financial trading, regulatory risk prediction, product recommendation engines, marketing personalization, offer management, data & app security, industrial IoT, smart cars, and more demand in-event intelligent decision making. Latency of mere milliseconds in these use cases could lead to not just financial losses of millions of dollars, but even loss of life in some cases.