A Data Science Central Community
I've been deliberately provocative in the title here, but while I'm relatively green in the analytical world, I am genuinely struggling with this question. I'm hoping to justify time spent understanding and implementing Association Rules at a retailer for which I currently work.
My understanding is that Association Rules are useful in understanding genuine affinities between products, in the form SKU A => SKU B (where A might be an itemset rather than a single SKU). However, a network graph (eg. using Gephi) describing relationships between items in a basket would also demonstrate such affinities - providing a helpful metric has been defined - and does not require filtering of trivial or non-interesting results. Furthermore, it harnesses the power of our visual system to understand clusters of items rather than single rules. We can then in conjunction with subject matter experts understand quickly what the interesting rules are, but also define wider patterns and understand aggregate customer behaviour.
Anyone willing to shed some light on the incremental benefit of using Association Rules?