This exercise was proposed as a hackaton by Analytics Vidhya, a community of data science enthusiasts.
The idea was to predict Purchase Amount from a dataset containing some user usage information, like Product ID and Product Categories, and demograpic information, like Gender, Age Range and Time Living in the City. I did work on the assignment as requested by the hackaton, but my main aim here was to explore the tuning process and try to derive some first conclusions from it. I was particularly interested into checking how many iterations the model needed to run to reach a stable conclusion of which parameter value was the best, and to check the degree to which such a value was indeed impactful in the model performance, comparing to other values close to it. These elucidations could easily convert into time saving in the modelling step of such a Data Science project.
I considered this to be an important first attempt, from my part, to better understand the dynamics of tuning.