As we continue our series on formalizing a QA process surrounding data analytics, we wanted to address the very critical step of determining the appropriate filters to narrow down a data set. In the hypothetical case we discussed, regarding a online music radio service trying to determine their most “popular” music, selecting the appropriate filters is very challenging. Numerous approaches can be taken to get the most “popular” music including most played track, most played artists, most “+1” or “Like’s”, most searched for songs or artists, and the list goes on. So, it is important to take a step back when approaching such a project and discuss with the stakeholder to ascertain the appropriate filters. Often, a mix of different factors can be used to determine a weighted average popularity score. Nevertheless, it is important to reach a consensus as to which factors should be weighted heavily versus those that are less important.
Such a situation can easily get messy if proper planning is not done at the start of the project.
No comments:
Post a Comment