As we reach the midpoint of 2021, we’re thrilled to announce our second major release of the year! In AutoML, automated feature discovery with push-down integration for Snowflake is now generally available. Automated Time Series now includes automatic data prep, and we have enhanced our unique Eureqa forecasting models. MLOps users can now perform major lifecycle operations on remote models using our new MLOps management agents. We’ve also introduced No-Code AI Apps that allows you to quickly create beautiful and powerful AI applications using a visual drag-and-drop user interface. No coding skills required.
And this is just a fraction of the release. Are you excited? So are we, so let’s jump in.
Automated AI Reports
Trusted Insights for Your AI Projects. Automated AI Reports are now available. AI reports are designed to summarize the most important findings of your modeling project to stakeholders in an easily consumable format. In just a few clicks you receive a comprehensive summary of your AI project. The report provides accuracy insights for the top-performing model, including speed and cross-validation scores. It also captures interpretability insights from the Feature Impact histogram for your top-performing model. Detailed model explanations , performance metrics, and ethics insights generated in the AI Report help you build overall trust in your AI projects and prove value to your key stakeholders.
Feature Discovery Push Down Integration for Snowflake
Embrace The Data Cloud. In DataRobot AutoML, feature discovery push down integration for Snowflake is now generally available. Adopting AI across your organization has significant hurdles. DataRobot’s Feature Discovery streamlines this process by offering automated feature engineering, enabling the discovery and creation of valuable new features for your machine learning models from multiple related datasets. This capability is now tightly integrated with Snowflake to ensure all processing is pushed down onto the Data Cloud making the computation of new model features faster, more accurate and more cost-effective. Joint Snowflake and DataRobot customers can now benefit from this advanced offering that automates one of the most critical and important tasks in data science.
Feature Discovery Relationship Quality Assessment
Have Confidence to Relate Your Data. In Release 7.1, we have introduced a new relationship quality assessment capability for Feature Discovery in DataRobot AutoML. This feature allows you to understand the overall quality and detect potential problems when defining relationships between primary and secondary datasets, before starting the feature engineering process. This reduces the risk of wrong or bad relationships earlier in the modeling process. The relationship quality assessment automatically assesses likely join percentage, detects missing data, and incorrect feature derivation windows and provides guidance, so you can address relationship quality issues right away.
Automated Data Prep for Time Series
Solve the Most Common and Complex Time Series Data Preparation Tasks Automatically. DataRobot Automated Time Series now offers automatic data prep capabilities, solving the most common issues with time series datasets, including gap handling and dataset aggregation. If your data contains gaps, DataRobot will identify the problem and suggest a way to fill the gaps in just a couple of clicks. You can also aggregate your data to fit your forecasting needs. For example, you can turn transactional data into a daily summary or a daily summary into a weekly or monthly summary. DataRobot also exposes the underlying Spark SQL allowing experts to edit query as desired. Designed for both expert and novice time series users, automatic data prep gets you over the most tricky hurdles with no interruption to the modeling process.
Nowcasting for Time-Aware Models
Build a Better Understanding of the Present. Nowcasting is a modeling approach that can help many organizations collect critical insights by estimating the present, as yet unknown, conditions of the target variable. For example, these could be current-day pricing insights or recent changes in customer behavior.
In 7.1, we created a unique user experience for nowcasting projects. You can now access a much more comprehensive range of blueprints and use additional time-series settings to configure your modeling process. Nowcasting also supports a blind history gap in project settings to help you to improve imputation and address usual reporting delays. Additionally, built-in explainability tools help you read and understand strategic insights with ease.
Time Series Eureqa Model Enhancements
Our Revolutionary Eureqa Models on Autopilot. In Release 7.1, DataRobot Automated Time Series now runs its unique Eureqa forecasting models as part of the regular Autopilot process. DataRobot’s powerful Eureqa models are based on the idea that a genetic algorithm can fit different analytic expressions to trained data and return a mathematical formula as a machine learning model. Eureqa models return human-readable analytical expressions, providing maximum transparency for subject-matter experts to review and understand. It’s a fundamentally different approach compared to traditional supervised machine learning models such as tree-based, regression, or deep learning. Our approach has been cited in over 800 peer-reviewed publications and is already helping organizations in use cases ranging from finance to neuroscience.
Prediction Jobs and Scheduler
Manage and Maintain Your Prediction Schedules in One Place. In Release 7.1 of DataRobot MLOps, you can now easily set up and configure all prediction jobs for production deployments from a single place in the UI. We have dramatically simplified the user experience for you to create, manage, and monitor prediction jobs as an individual or globally as an MLOps administrator. You can also see job history and monitor job status directly in the MLOps user interface and access information on how your resources are being used and by whom. Prediction scheduling allows MLOps administrators to maintain predictions as a continuous process without needing to state the same job specifications every time.
MLOps Management Agents for Remote Models
Advanced Lifecycle Management for Any Model. Management agents are new to MLOps in release 7.1. Management agents are designed to understand the state of your remote models and can automate tasks including the retrieval of model artifacts and the deployment or replacement of your production models directly in their remote environment. They provide extensible support for a wide variety of use cases and pair easily with existing MLOps Agents to provide comprehensive monitoring and lifecycle management for any production model regardless of how it was created or where it is running.
No-Code AI Apps
Unlock the Business Value of Your AI Investments. We are excited to introduce No-Code AI Apps, which allows you to quickly turn any deployed model into a rich AI application without a single line of code. Using pre-built templates and drag and drop widgets, you can quickly and easily configure your app in the precise way you need to support your use case. AI Apps can be built to help decision makers to score new data, perform what-if scenarios, and even run hundreds of simulations to identify the ideal combination of input values to optimize the target outcome. Built-in explainability tools help users better understand the predictions made to justify the decision-making process. AI apps also allow you to iterate and improve your models through constant feedback from the business community.
More Coming Soon. Stay Tuned!
Of course there is way more to release 7.1 than the items listed above, so be sure to check out the product release center for a full list of new and improved features. At our recent AI Experience 2021 event we also introduced Composable ML, Bias and Fairness Monitoring, and Continuous AI. These groundbreaking product enhancements are coming later this year and will further strengthen our platform and enhance value for our customers.