---

Mining Association Rules

---

• Association Rules
  • Association rules show attribute value conditions that occur frequently together in a given dataset.

  • Example Association Rule:

    98% of customers who purchase tires and auto accesories also get automotive services done on their car.

    ---

  • A widely-used example of association rule mining is Market Basket Analysis:

    Market Basket Analysis is a modelling technique based upon the theory that: IF you buy a certain group of items, THEN you are more (or less) likely to buy another group of items.

    ---

  • Association rules provide information of this type in the form of "if-then" statements.

    The IF part of the association rules is called the antecedent Example: IF you buy a certain group of items The ELSE part of the association rules is called the consequent Example: THEN you are more (or less) likely to buy another group of items

    ---

    ---

  • Association rules are always probabilistic in nature.

  • An association rule has two numbers that express the degree of uncertainty/confidence about the rule:

    Support Confidence

    ---

    ---

  • Support

    The support is the number of transactions that include all items in the antecedent and consequent parts of the rule. The support is sometimes expressed as a percentage of the total number of records in the database.

    Example:

    Transactions: {coffee, milk}, {coffee, milk, candy}, {coffee, milk, juice}, {coffee, candy}, {coffee, milk}, {milk, candy} Association rule: if you buy coffee, then you will also buy milk Support{coffee, milk} = 4

    ---

    ---

  • Confidence

    Confidence is the ratio of the number of transactions that contains all items in the consequent as well as the antecedent (i.e, the support) to the number of transactions that contains all items in the antecedent.

    Example:

    Transactions: {coffee, milk}, {coffee, milk, candy}, {coffee, milk, juice}, {coffee, candy}, {coffee, milk}, {milk, candy} Association rule: if you buy coffee, then you will also buy milk Confidence{coffee, milk} = 4/5 = 80% In other words: In 80% of the time, when you buy coffee, then you will also buy milk

  ---

  ---

  ---

  ---

• Data Mining
  • The association rules are computed from the sales or other types of data

    The process of extracting informating from data is called data mining

    (In mining, 99.9% of the time you get garbage (dirt) and rarely you find something valuable.

    This is also the case in data "mining")

    ---

  • Data mining is the process of extracting patterns from data.

  • As more data are gathered, data mining is becoming increasingly important.

  • Usages of data mining:

    Profiling consumer behavior Profiling in surveillance (what is the normal pattern, so breach of sucurity can be detected) fraud detection and even scientific discovery...

  ---

  ---

• Data Mining techniques used in extracting information from data
  • Data mining extracts information through the following classes of data analysis techniques:

    Classification techniques: Arranges the data into predefined groups. Example algorithms: Nearest neighbor, Naive Bayes classifier and Neural network. Clustering techniques: the algorithm will try to group similar items together. Example algorithms: k-means clustering Hierarchical clustering Regression: find a function which models the data with the least error. Association rule learning: finds relationships between variables. Example algorithms: A-priori Frequent Pattern-growth algorithm For example a supermarket might gather data of what each customer buys. Using association rule learning, the supermarket can work out what products are frequently bought together, which is useful for marketing purposes. This is sometimes referred to as "market basket analysis".

  ---

  ---

• Association Rules Mining
  • We will first study association rules mining

  • Definitions:

    U = universe of all items X = a set of items (a subset of U) X is also called a itemset T = set of transactions T is a set of itemsets The support for X (support(X)) = fraction of transactions in T that is a superset of X (i.e., that contains all items of X). The confidence for a rule X => Y is defined as: Confidence(X => Y) = Support(X ∪ Y)   /   Support(X)

    ---

    ---

  • The problem of mining association rules is:

    Generate all association rules have: support greater the user specified minumum support confidence greater the user specified minumum confidence

  • The problem of discovering all association rules can be decomposed into 2 sub-problems:

    Find all sets of items (itemsets) that have transaction support above the user specified minimum level The itemsets with sufficient support are called large or frequent itemsets (heavy elements) The itemsets with less than the minimum support are called small or not frequent itemsets (light elements) Use the large itemsets to generate the association rules The general idea in generating association rules is: Find large itemsets A ∪ B and A (We have satisfied the minimum support requirement) Calculate: Confidence(A ⇒ B)   =   Support(A ∪ B)   /   Support(A) If Confidence(A ⇒ B) exceeds the minimum confidence level, we have found an association rule !

    ---

    ---

  • The first step to perform association rule mining is to find all itemsets that have sufficient support

    I.e.:

    Find all X such that: Support(X)   ≥   θN

    ---

  • This is the problem of find heavy elements.

    Except now the elements are itemsets (instead of simple values)

  ---