Build higher performing and more accurate predictive models with new DataRobot features

While generative AI is dominating the headlines, the reality is that the majority of AI use cases that drive measurable business value today are predictive use cases.

We recently launched 22 new features designed to help you scale predictive AI solutions and ensure model integrity and performance from build through deployment.  

Today, we’ll explore some of the new enhancements that allow you to quickly prepare data for modeling and evaluate model performance when building predictive AI models in DataRobot. 

💡Pro tip: Build customized projects that harness the combined power of predictive AI and generative AI with DataRobot for new levels of innovation and impact. 

Enhancing AI Data Prep for Model Accuracy and Performance 

Few steps are as tedious as transforming and preparing data for modeling. At DataRobot, we’ve always made it easier to get your data AI-ready, even dirty data, which we handle for you with ease. Using Datarobot means that you never need to drag-and-drop data prep before you model, you just need to point DataRobot at a file or table and let the platform do the rest. We’ve now added all of the great functionality you know and love about our AI data auto-prep from our Classic UX to our new NextGen interface.  

Secure Data Connectivity: Find, share, and leverage data easily with enhanced browsing and preview functionality, profile details, in cloud data warehouses, cloud storage, and the AI Catalog in NextGen.

Wrangle, Join, and Aggregate: Enhance your data workflows by seamlessly joining, aggregating, and transforming data directly on supported cloud data warehouses or data stored in the DataRobot AI Catalog and blob storage. Point DataRobot to one table (or several) and quickly identify if there is any signal in your data, then easily materialize this data into your data warehouse for reuse in NextGen. 

Feature Discovery: DataRobot has always been unique in how we perform feature engineering and feature discovery. You can now access all these rich features and build recipes for your specific use cases to generate new datasets with derived features in NexGen. 

💡Pro tip: If you’re on the SaaS version of DataRobot, you already have access to these new features in the latest version of DataRobot. If your organization uses our on-prem solution, you’ll need to manually update DataRobot to see our latest and greatest enhancements.

AI-Driven Insights and Explainability At Your Fingertips

Explainability is essential for building trust in your models. Whether you’re looking to deliver an AI-driven recommendation or making the case for the productionalization of a model, being able to interpret how a model works and makes decisions is a critical capability. 

Not only is explainability essential for gaining adoption of your models from business stakeholders, it’s also important in helping you understand the key drivers of outcomes and gain deep AI-driven insights. A clear understanding of the how and why your models work enables you to create stronger change within your organization. We’ve extended and added more of these insights into our NextGen UX. 

Explain Predictions with SHAP Insights: Quickly understand predictions with enhanced SHAP explanations support for all model types and new individual PE functionality that calculates SHAP values for each individual row.

Slices Insights: Enhance your understanding of how models perform on different subpopulations by viewing and comparing insights based on segments of your project data. Slice data by date/time, numerical, categorical, and boolean data types. 

Easily Compare and Optimize Models 

Our newest features included in Workbench make it easier than ever to train and compare different predictive models in DataRobot. Not only can you quickly select between experiments and evaluate key performance metrics, we’ve now incorporated new insights into the NextGen UI that enable you to quickly understand model effectiveness and improve performance. We’ve also begun the process of moving over all of the multimodel capabilities we offer in our Classic UX to NextGen, starting with Time Series: 

Enhanced Confusion Matrix: Train classifiers on datasets with unlimited classes within Workbench, then quickly understand the effectiveness of your classifiers with our enhanced confusion matrix.

Side-by-Side Modeling Insights: Rapidly improve model performance by easily assessing model performance and comparing models across experiments, even those that use varied datasets and modeling parameters.

Time Series Experience: Easily build robust, fine-grained time series forecasts in our new NextGen UX and explore the new functionality we’ve added.

A Unified View Across Notebook and Non-Notebook Files 

For our code-first users, we have invested significant resources in giving you a best-in-class experience. In this release, we enhanced our codespaces to allow you to focus on building models, not infrastructure, by opening, viewing, and editing multiple notebook and non-notebook files simultaneously. New enhancements make it even easier to edit and execute files, as well as develop new workflows. 

​​Codespaces and Codespace Scheduling: Build reusable automated workflows with new Codespace features. Open, view, edit, and execute multiple notebook and non-notebook files in the same container session. Easily establish automated jobs at any desired cadence. Monitor your scheduled notebook jobs and track run history. Configure scheduled notebooks to develop automated, reusable workflows for effortless execution.

Near-Infinite Scale at Modeling and at Inference Time  

Data is exploding, leading to a massive increase in the data sizes with which teams are working on a daily basis. With this new release, we’re not just giving you the ability to work with larger datasets at build and inference time, we’re doing so in a hyper-efficient way. 

Constantly increasing cloud costs are beginning to pose a major challenge to AI teams, who need to balance effective training with budget constraints. Since our founding in 2012, DataRobot has been focused on helping data science teams maximize their investment. In this case, we do so by not charging on a consumption basis, unlike most AI and data platforms, which are motivated to increase your cloud costs. Our latest release further increases the value of your hard work by allowing your team to freely work with big data without worrying about costs. 

Scale Enhancements: Seamless handling of large datasets throughout the ML lifecycle with incremental learning and enhanced NVIDIA GPU compatibility. Our incremental learning is designed to get you to the best model, not just chug through processing all your data. It will also alert you when you get diminishing returns on using more data, so you’re not wasting time when modeling. 

💡Pro tip: Easily move projects and datasets into the latest DataRobot experience with expanded Project Migration features to take full advantage of all of the new functionality, visuals, and collaboration features.

Features Designed to Deliver Impact

Though GenAI is consuming a great deal of attention, we know that many of you are seeing significant success with predictive AI. Our latest launch showcases how DataRobot is continuing to invest in predictive AI, while many other AI vendors are chasing the hype cycle and sidelining their predictive AI products. We know that true impact requires a combination of predictive AND generative, and DataRobot is where AI teams turn to to deliver tangible results for their business.  

Our customer community continues to uncover new use cases and mature existing AI initiatives with incredible momentum: the average projects per customer have increased 12% in the past year while predictions have increased 11% per customer. 

With the latest DataRobot enhancements, you have greater control over critical early development stages. But the innovations don’t stop there. Stay tuned for further deep dives into our Summer Launch ‘24 as we explore recently introduced features that streamline how you deploy, observe, and manage your predictive models.

About the author

Lisa Aguilar

VP, Product Marketing, DataRobot

Lisa Aguilar is VP of Product Marketing and Field CTOs at DataRobot, where she is responsible for building and executing the go-to-market strategy for their AI-driven forecasting product line. As part of her role, she partners closely with the product management and development teams to identify key solutions that can address the needs of retailers, manufacturers, and financial service providers with AI. Prior to DataRobot, Lisa was at ThoughtSpot, the leader in Search and AI-Driven Analytics.

Meet Lisa Aguilar