This task is to take some of the algorithms and setup a test suite, that show performance of the system with different properties of the input data set.
There are two main approaches:
select common data set and train to competitive accuracy.
generate data, with specific properties and run algorithms a fixed set of iterations.
The entire process from the downloading of data set, to preprocessing before the algorithm is run have to be automated to the point of a single script being executed, and resulting in performance numbers. If preprocessing is required, then bench marking this is also valuable.