Choosing a machine learning algorithm depends on the task (regression, classification, clustering, and so on), the size and shape of the data, interpretability requirements, and compute constraints. In practice, simple baselines like linear models and tree ensembles are worth trying first before reaching for more complex methods.