BNet.Builder Provides Unprecedented Ease-of-Use
Charles River Analytics today announced the launch of its new Bayesian belief network modeling software, BNet.Builder™. This desktop software offers a graphical user interface with unprecedented ease-of-use, and a unique Dynamic Update© feature that allows the visualization of changing causal relationships by showing users how changing the network affects beliefs in real time. BNet.EngineKit™, a developer API, will be released in mid-2005.
BNet.Builder provides ease-of-use and visualization capabilities unique in the market. Dynamic Update allows the user to see beliefs change in real-time in response to modifying any aspect of the network, including conditional probabilities and evidence, without being forced to switch modes or re-compile. A number of features make the construction of Conditional Probability Tables easier and faster, including drag-and-drop column reordering, cut-and-paste compatibility with common spreadsheet applications, and setting the values of multiple rows at once. Further ease-of-use comes with the ability to copy and paste nodes and links complete with their probability tables. Written entirely in Java, BNet.Builder is available for Windows, Mac OSX, and Linux operating systems. All of these features will also be available in BNet.EngineKit.
BNet.Builder was originally developed as a tool for the scientists and software engineers at Charles River Analytics when they were unable to find existing software that met their needs. Greg Zacharias, President of Charles River Analytics noted, “We developed BNet.Builder for our intelligent systems product development and use it extensively in many of our projects. As a result, it reflects the right balance of functionality, usability, and display visualization needed for the wide range of applications the general R&D community faces all the time.”
Karen Harper, Vice President and Principal Scientist at Charles River Analytics, was in charge of a project that drove the early development of the tool. She noted: “We had used other belief network software, but we found that none offered the right set of features and the polished user interface that the project required. Other applications were unintuituve and too clumsy for entering data gathered from knowledge elicitation sessions with experts. The available APIs provided limited access to underlying models and their GUI components were too awkward to embed effectively in our research applications. BNet has offered us an orders-of-magnitude improvement in usability and software development capability. Working with the prototype version of EngineKit has sped the integration of belief networks in our intelligent agent toolkit.”