Using rates of arrival to identify change points

Another strategy involves analyzing the rate of arrival of data points to spot changes. This subtle method targets less obvious shifts in the data, which may not exactly qualify as anomalies, but represent significant alterations nonetheless. The underlying premise is that changes in the rate of arrival likely indicate a shift in the data-generating process. This technique proves particularly useful in fields such as sales, where a sudden drop in the rate of sales can signal a change in the data-generating process. To model this, we consider the sales data as a Poisson distributed process, and the rate of arrival of sales as an exponential distribution.

A Poisson distribution is a type of probability distribution used to calculate the likelihood of a certain number of events occurring within a specified time or space interval. This calculation assumes that the events take place at a constant average rate and that each event happens...