It's simple: first, define the problem. Unbelievably, that's often the hardest step. Then an evaluation metric depending on your business goals and data imbalance. In the figure above at the left, you prepare your data (ideally, or we help you with it as well). At the right, we identify and provide the algorithms we believe is the best to solve your problem. Then the data meets the algorithm and that makes for a hot mix.
But how we choose our models? There are hudrends and thounsands of different possible algorithms to use out there. It's the same for ways to preprocess your data into the algorithm. That requires expertise, creativity and investigation. It also entirely depends on the identified problem to solve, and your quantities of data as follow: