This was a team project for my Business Analytics course in university, where I worked in a group to identify a dataset to analyse for the sake of providing a hypothetical benefit to a hypothetical business which we would then pitch to the lecturer at the end of the term.

As a group, we came up the idea for analyzing an extensive video game sales database to try and find the optimal type of game to maximise video game sales in different regions of the world.

To approach this task, we split up into groups to, process/clean the data, supplement missing data through web-scraping, remove outliers, and apply a multitude of different supervised and unsupervised algorithms on the dataset.

My focus was on outlier removal as well as K-means clustering for an unsupervised approach to see any patterns that could be algorithmically spotted in the data. I used the elbow-method for determining the best amount of centroids to cluster around, which allowed for useful insights into how different video games were more popular in different regions of the world.

There are no hard an fast rules to truly determine the perfect parameters for an algorithm to use when analysing data. It is very useful to create graphical representations of your findings along the way so that you can tune parameters for anything the algorithms have missed.