Piecewise regression: when one line simply isn’t enough

Piecewise regression is a technique used when a single line isn't sufficient to model data. It breaks the domain into segments and fits a separate line through each one. Datadog has automated this process for time series data, with objectives including automated breakpoint detection and segment count detection. The algorithm uses a greedy approach to find the best solution, with stopping criteria based on error increase. Examples are provided to demonstrate its effectiveness. The code is available in the Datadog/piecewise Github repo.