Now that we know about the basics of Bayes' rule, let's try to understand the concept of Bayesian inference or modeling.
As we know, real-world environments are always dynamic, noisy, observation costly, and time-sensitive. When business decisions are based on forecasting in these environments, we want to not only produce better forecasts, but also quantify the uncertainty in these forecasts. For this reason, the theory of Bayesian inferences is extremely handy as it provides a principled approach to such problems.
For a typical time series model, we effectively carry out curve fitting based on y when given the x variable. This helps to fit a curve based on past observations. Let's try to understand its limitations. Consider the following example of temperature in a city:
Day | Temperature |
May 1 10 AM | 10.5 degrees Celsius |
May 15 10 AM | 17.5 degrees Celsius |
May 30 10 AM | 25 degrees Celsius |
Using curve fitting, we obtain the following model:
However, this will imply that the...