Association Rule Mining

• ARM (Association Rule Mining) is an UnSupervised ML Model that’s used to find the associations present between different instances/objects.
  • Its widely used in the Retail Industry for Market Basket Analysis (MBA) i.e. analyzing the shopping carts of the customers to find the items that are purchased together to build the rule of associations.
  • It can also be used in Healthcare domain to understand association between various pathogens.
  • It can be used in Chemistry domain to find associations between different elements/chemicals.
  • It can also be used in Telecommunication and Weather Department.

• Algorithms used for ARM modeling are :
  1. Apriori Algo
     • Given by R. Agrawal and R. Srikant (1994) for finding frequent itemsets in a dataset for boolean association rule. 
     • Known as Apriori because it uses prior knowledge of frequent itemset properties.
  2. ECLAT
     • Equivalence Class Clustering and bottom-up Lattice Traversal
     • It is a more efficient and scalable version of the Apriori
  3. FP-growth
     • Frequent-Pattern growth, internally uses FP-tree (Frequent Pattern tree)
     • Improved version of the Apriori Algorithm
     • Performs better than ECLAT in most cases

Apriori Algorithm is the widely used algo in this field.

It helps in identifying which items/products are frequently bought together by customers.

Below picture depicts important terms/formulas used in Apriori Algorithm. 
    1. Support
    2. Confidence
    3. Lift

Its showing an Association Rule between Item X and Item Y, to understand/calculate that if Someone buys Item X then what are the chances/probability that he/she will also buy Item Y.

Rules must be useful, but not trivial or Unexplicable.