English 
dotData, a leading data analytics platform, announced product updates for its dotData Feature Factory 1.4 and dotData Ops 1.5 on the 20th. These updates include significant enhancements, including improved data quality, improved output interpretability, and model retraining, enabling more reliable AI to be built and deployed faster than ever before. dotData Feature Factory 1.4 now supports direct deployment of dotData Feature Factory on Microsoft Fabric. Microsoft Azure users can automatically generate high-quality features on Microsoft Fabric without moving or replicating data. With centralized data governance on Microsoft Fabric, dotData Feature Factory can be operated with scalability, security, and flexibility.
Additionally, the platform has enhanced AI-based data checking and cleansing. AI automatically detects a wide range of data quality issues, including constant columns, mixed category codes in numeric fields, and values outside of expected formats. Corrective actions are suggested based on the detected results and can be immediately applied from the UI to proactively resolve potential data issues. dotData Ops 1.5 replaces tedious and challenging data cleansing with an interactive, AI-assisted process, improving the reliability of your analysis.
In addition, widgets such as feature leaderboards can now be exported as standalone HTML. Features and analysis results can be quickly shared with business departments without access to an analytics environment, allowing them to interactively review and interpret results rather than using static reports. This speeds up the business feedback loop and accelerates the discovery of valuable patterns. dotData Ops 1.5 includes a framework for comparing and managing deployed “champion” models with new “challenger” models. Challengers can be created by uploading existing models or by automatically retraining them using AutoML (automated machine learning) with the latest data. Automatically searching for the best model candidate at any time streamlines addressing accuracy degradation. It also supports automatic switching between shadow testing and production models. Run a challenger model in parallel with the champion model running in production to compare and monitor prediction accuracy and feature drift without impacting the production environment.
こちらもお読みください： Tech Data and HCLSoftware Expand APAC Partnership
For example, by setting criteria for automatically switching production models, such as “the challenger outperforms the champion in three of the last four tests,” the model will be automatically adopted for production when the criteria are met, ensuring that predictions are always provided using the optimal model. dotData Feature Factory 1.4 and ドットデータ Ops 1.5 also include various improvements in robustness, ease of use, and automation for managing the entire machine learning model lifecycle.
ソース ヤフー
日本語