Machine Learning with Sesam
Solving Data Integration for Machine Learning
Machine learning (ML) is limited by the quality and amount of relevant data readily available for modelling and scoring. The required data preparation effort is time consuming and expensive, quickly consuming allocated project resources. Not having enough data, not having the right data or even having too much of the wrong data is proven to be a significant barrier to delivering tangible business value when adopting of ML.
Sesam’s unique data driven architecture enables the rapid uptake of ML in the organization by quickly and simply facilitating the required data and systems integrations.
Sesam provides operational patterns for both training and operating ML models. With the increased agility provided by Sesam’s in built tools, data from all your enterprise systems can be powerfully combined, shaped and transformed into rich contextual feature sets and quickly delivered to directly modelling and scoring algorithms. This reduces the time and complexity usually required to deliver relevant business value.
- Mature data ecosystem operationalizes Machine Learning into your organization
- Rapidly reduce start-up costs in sourcing and delivering clean data into ML projects. 80% of the project time used on data acquisition can be reduced to 50%. Reusability multiplies these gains with ongoing agility over time
- Immediately propagate ML results to all systems via data driven architecture
- Low cost start up – offered as a cloud service (iPaaS)
- Easily combine, shape and iterate on rich contextual data to create feature sets
- Easy and consistent data acquisition with standard source connectors
- Quickly leverage data from all your systems, including service analytics
- Reduced license and operational costs from previous bespoke systems
Sesam Machine Learning Architecture
Here we integrate Sesam with Azure Machine Learning Studio
- 1. Sesam is able to powerfully merge, combine and clean data entities from all your enterprise systems. This creates powerful and reusable global datasets, with strong name spacing, management of context and mapping of ontologies.
- 2. From these global datasets, ML feature sets can be defined, shaped, and transformed ready for consumption by training score models and ML algorithms within Azure Machine Learning Studio.
- 3. Iterations of the ML feature sets are easy to create and recreate in Sesam, as requirements evolve whilst data scientists work on modelling tasks.
- 4. Once a ML model has been created and tested, it can be loaded into Azure ML Webservices within the Azure Machine Learning Studio.
- 5. Sesam pipes are configured to supply the same feature set data definition used by the training and scoring process to the operational webservice. This enables both the initial load across all existing data, and thereafter processing the incremental and ongoing additional additions and changes to that data set.
- 6. The data results generated by the ML processing is returned to the datahub. It is both merged with the global dataset enriching the context, and also can be simply propagated further to any system within your enterprise.
Last edited: 03.09.2019 | Published: 03.09.2019
All your data – Connected