Improving models and predictions
My previous post illustrated modeling through the SIR epidemiological model. This one describes how models are updated based on how well their predictions aligned with real-world outcomes. These include updating variables, settings, and algorithms to improve predictive accuracy.
For example, the SIR model's beta (β) variable is the product of the average number of contacts a susceptible person makes per day (c) and the probability that a susceptible person will become infected after each contact with an infected person (p), β = p * c. There are β values for common infectious diseases like measles based on case data and analysis by entities like the USCDC and the WHO. β values are estimates in progress for new diseases and variants, including COVID-19.
I've have worked on many models that relied on transactional data in areas like accounting, safety, and quality control. Some benefitted from existing standards, while others shed light on poorly quantified areas of performance. In contexts like homeland security, models and analytics helped us identify conflicts in tactics, policy, and legal constraints. AI is adding new tools that are revolutionizing these practices, that's the subject of the next post.