According to Forrester Research’s 2018 IIoT Wave, released August 9, the C3 Platform is the highest scored offering (tied) for advanced data analytics at enterprise scale.
The C3 Platform has been purpose-built to support advanced analytics on large data sets unified from disparate source systems. We realized early on that organizations seeking to capture value from industrial IoT need to place algorithms and models at the center of their business operations. Therefore, in developing our advanced analytic capabilities, we particularly focused on streamlining the development, deployment, provisioning, and operation of machine learning and AI models at large-scale, in real-time, and in production.
C3 offers the following key benefits to data scientists seeking to place their models into production:
C3’s unique, model-driven AI architecture enables data scientists to access data in a unified, federated data image, regardless of the underlying data store. Cross-data store queries (e.g., across key-value, relational, or file systems) are significantly streamlined. Key concepts of time and space, data normalizing, versioning, and tracking are all handled by the C3 Platform. So data scientists can focus on functional performance.
Data scientists often spent 80-90% of their time massaging data into data frames for model training. The C3 Platform and the C3 Type System shrinks data scientists’ data preparation work, and enable them to build models that can readily transition into production.
- The C3 Platform enables integrated algorithm development in common data science tools (e.g., Python, R) with bindings to the C3 Type System, and via the use of visual tools. The C3 platform supports continuous training of algorithms based on field feedback. Models are deployed rapidly and are available via REST APIs or embeddable in applications – passed data, hosted, and executed using the C3 Platform.
- C3 offers native support for a comprehensive set of machine learning frameworks to facilitate rapid development of AI-enabled applications. Data scientists have access to standard ML libraries, C3 libraries and tools, and the full set of data—eliminating the need to extract subsets of data into third-party tools. Data scientists are also able to readily specify the libraries and runtimes required for executing their algorithms. The C3 Platform ensures the same libraries and runtimes are available in development and production.
- Data scientists are able to build sophisticated pipelines by using the “scaffolding” provided through C3 Pipeline Types. These types enable data scientists to train and deploy models, including pre-and post-processing steps as complex, composable graphs without requiring custom coding for production use.
- Using the C3 Type System, data scientists are readily able to manage and version their models. They are able to track the performance of models and enable models to be retrained (or redesigned) if performance starts to drift, or additional data arrive. Models can be readily deployed and promoted from stage, to QA, to production environments.
- The C3 Platform has native support for parallelized batch, stream, and—most importantly—continuous processing frameworks. These frameworks enable analytic models to be executed in flexible ways. Continuous processing is often required for models that require processing as new data arrive, but also require deep-contextual lookup (most real-world enterprise machine learning/AI use cases follow this pattern). C3’s continuous processing engine allows for selectively processing data pathways that require re-processing based on incremental data changes. This enables scores/alerts to be kept current in near-real time as new data flow in, while minimizing computational cost.
All of these elements together demonstrate why C3 offers the most powerful and usable advanced data analytics, an essential component of the rapid development of industrial IoT and digital transformation efforts today.