Gráfico de forecasting

Here are some insights from different podcasts discussing the use of graphs in forecasting:

1. Traffic Forecasting Using Graph Neural Networks:

   • Eric Manibardo discusses how graph neural networks can effectively model entire traffic networks for forecasting. By using graphs, these models can predict traffic states across a city, improving performance by understanding traffic correlations between road segments. This approach requires less data specificity for each street, making it more scalable for city-wide implementation 1.

2. Weather Forecasting with Machine Learning:

   • David Friedberg highlights a machine learning model that outperforms traditional methods in weather forecasting. Using graph neural networks, this model allows for faster and more accurate forecasts, even for extreme weather events. The ability to run sophisticated models on smaller computing devices reduces the reliance on massive computations traditionally used in weather predictions 2 3.
     • 

       1

       Traffic Forecasting Insights

       00:22:57
     • 

       2

       Revolutionary Weather Forecasting

       01:49:27
     • 

       3

       Weather Forecasting Revolution

       01:52:02
     • 

       4

       Improving Probabilistic Forecasts

       00:31:23

3. Calibration in Forecasting Models:

   • Michael Mauboussin discusses the importance of calibration in probabilistic forecasting. By plotting subjective probabilities against actual outcomes, forecasters can refine their models over time. This process, similar to practices by weather forecasters, helps improve the accuracy of predictions and can be done using simple tools like spreadsheets 4.

These examples show the diverse application of graph-based methods in forecasting across different domains, showcasing their effectiveness in handling complex, multi-variable systems.