A new machine learning product that makes use of Amazon SageMaker has been announced by Periscope Data. The system is currently in beta.
Periscope Data is best known for its Periscope Pro product which can be used to analyze data in SQL, Python or R on the same development environment. The software can then be used to visualize the data, collaborate and share with others. The software includes predictive analytics, machine learning, natural language processing and data cleaning.
Amazon SageMaker is a fully-managed platform that can be used to build, train, and deploy machine learning models at any scale. It includes hosted Jupyter notebooks for exploring and visualizing your training data stored in Amazon S3. You can connect directly to data in S3, or use AWS Glue to move data from Amazon RDS, Amazon DynamoDB, and Amazon Redshift into S3 for analysis in your notebook.
The Periscope Data ML system aims to combine all the stages of creating a working model, from data preparation through model training to reporting. The developers say that this will overcome the problems of having fragmented workflows when building ML models involving manual steps and disparate tools. In their experience, data preparation and feature engineering are particularly difficult because of the iterative experimentation involved in the ML process.
The new system starts with using Periscope Data to make the data preparation as simple as possible by surfacing distribution and descriptive statistics, and by helping users recognize where more data cleaning or normalization is required. Teams working on the data can also make use of SQL in views to create datasets, and to help work out what data features are relevant.
Once the datasets are created, they are directly exported into Amazon SageMaker to build, train and deploy ML models. Model results are then written back to a data warehouse, allowing users to use trained models for predictive and prescriptive reporting in Periscope Data.
The Periscope Data integration with Amazon SageMaker will be made available publicly in the coming months, or you can apply to join the beta program.