Presented by Dave Wells
Machine learning is at the cutting edge of modern data use cases. After years of slow and painstaking advances, machine learning is experiencing rapid adoption today. Machine learning uses statistical methods to make predictions and to automatically improve prediction accuracy over time. In business it is readily applied to common analytics problems such as quantitative investing, customer recommendations, medical diagnosis, predictive maintenance, fraud prevention, and more. And today’s hottest technologies – generative AI – are built on a machine learning foundation. We are in the early stages of a steep adoption curve for machine learning.
To prepare for machine learning, it is important to focus on Data Architecture. You don’t need (and don’t want) a separate Data Architecture for machine learning. Instead, think about how to extend your existing Data Architecture. What new Data Management capabilities do you need? What new architectural features and functions should be intergrated into your Data Architecture?
A Google data scientist once said that simple models working with very large datasets are more accurate than complex models with small amounts of data. So, data volume is certainly a consideration, but it is only the beginning. Join this session to learn about these Data Architecture considerations for machine learning:
- Data Architecture for massive data volumes
- Real-time data pipelines
- A single database for transactions and analytics
- Redeploying batch models in real time
- Data preparation for supervised learning
- Data preparation for unsupervised learning
- Write-back Data Management (machine learning creates new data)
