Scenario Unlocked unlocks solutions where human judgement and decision-making is involved. Sufficiently important decisions require deeper analysis. If a quantitative approach is available, it is usually the one preferred as it will offer the best combination of an approach and outcome.
Baseball AI Workbench is a web application that showcases performing quantitative decision analysis (decision thresholding, what-if analysis, AI Agents with probability & confidence interval analysis) using in-memory Machine Learning models with historical baseball data.
The application has the following features:
- Historical position player (batters) up to the end of the 2024 season
- Three different decision analysis mechanisms to perform what-if analysis
- Agentic AI integrations with Agents performing research & quantitative analysis
- A simple "expert" rules engine to predict baseball hall of fame induction, contrasted with a Machine Intelligence solution
- Single and multiple machine learning models working together to predict baseball hall of fame ballot and induction probabilities
- Machine Learning models are surfaced via ML.NET in-memory for rapid inference (predictions)
- Surfaced via the Aspire.NET integration with a Blazor application framework using SignalR to deliver the predictions from the server to the web client at scale
- Self-contained application with Docker, allowing you to run locally
Architecture - Cloud Deployment Diagram:
Project Structure (Verified):
- Visual Studio 2022, .NET 9, Server-Side Blazor, ML.NET v4.02, Semantic Kernel, Azure AI Foundy, Azure OpenAI Azure SignalR (optional for massively scaling message communication for Azure deployments)
More Information:
- ML.NET: https://dotnet.microsoft.com/apps/machinelearning-ai/ml-dotnet
- Blazor: https://dotnet.microsoft.com/apps/aspnet/web-apps/blazor
- Historical Baseball Statistics Database (used as the model training and inference data set): http://www.seanlahman.com/baseball-archive/statistics/
- How to Measure Anything (Amazon book link): https://www.amazon.com/How-Measure-Anything-Intangibles-Business-ebook/dp/B00INUYS2U/ref=sr_1_1?dchild=1&keywords=how+to+measure+anything&qid=1588713606&sr=8-1
- Decision Management Systems (Amazon book link): https://www.amazon.com/Decision-Management-Systems-Practical-Predictive/dp/0132884380