Association analysis measures the strength of co-occurrence between one item and another. There is a famous story describing association analysis known as the Beer and Diaper relationship with the idea being the customers who buy beer also tend to buy diapers.
In the BakerySales dataset, there are 435 transactions. The customer has the decision to buy from a variety of coffees, desserts, teas, and breads.
Executive Summary: The goal of analysis here is to find frequent item combinations given the transactional data. Support is a measure that counts the frequency of an item. Confidence measures the likelihood of an item given an outcome. Confidence is calculated by:
Confidence (X>Y) = Support(X U Y) / Support(X)
For my analysis I set SUPPORT to .2 and CONFIDENCE to .4. I felt these were reasonable settings to apply the algorithm. This leaves us with popular purchases and items that we can confidently say have an association together.
Table 1. Findings
As you can see those who buy coffee2 often buy Dessert1 and vice versa. It is one of the most popular combinations and I advise this bakery to feature it as a combo deal. In addition, the bakery could add another item to the combo, such as a bread. Furthermore, if the bakery has its goodies out on display, I’d recommend putting these items in close proximity.
Relating this algorithm back to baseball, I am not sure if there is any actionable insight for this in game data. My mind wandered around analyzing successful pitch combinations and reliever combinations. But that may be a stretch. Nonetheless, it could prove useful for concession and ticket sales.