#1 Solved Example Apriori Algorithm to find Strong Association Rules Data Mining Machine Learning

Summary

This video walks through the Apriori algorithm using a five-transaction dataset to generate strong association rules. With minimum support 40% (support count = 2) and minimum confidence 70%, the presenter computes support counts for 1-itemsets, selects frequent 1-itemsets, then builds and evaluates 2- and 3-itemsets to find frequent itemsets. After identifying the largest frequent itemset, the video derives candidate association rules and computes confidence for each direction (e.g., ocean → beach vs. beach → ocean), keeping only those meeting the 70% threshold. Clear step-by-step counts and examples make Apriori’s mechanics and rule selection easy to follow.

Takeaways

• 😀 The A Priori Algorithm is used to discover strong association rules in a given dataset based on predefined support and confidence thresholds.
• 😀 The dataset in the example contains five image IDs, each associated with a set of tags that will be analyzed using the algorithm.
• 😀 The first step in the A Priori algorithm is to calculate the support count for individual item sets (one-item sets) and write them down in a table.
• 😀 A frequent item set is determined by comparing the support count with the minimum support value, which, in this case, is 40%.
• 😀 After identifying the frequent one-item sets, the algorithm generates combinations of two-item sets and checks their support counts to find frequent two-item sets.
• 😀 The same process is applied to generate three-item sets by combining distinct two-item frequent sets and counting their support counts.
• 😀 Strong association rules are generated by calculating the confidence for each rule. The confidence formula is: (frequency of both X and Y) / (frequency of X).
• 😀 A strong rule is one where the confidence is greater than or equal to 70%. For example, ‘Ocean → Beach’ has a confidence of 100%, making it a strong rule.
• 😀 Rules such as ‘Beach → Sunshine’ and ‘Holiday → Sunshine’ also meet the minimum confidence threshold, making them strong rules.
• 😀 The process stops when no further frequent item sets can be generated (in this case, no four-item sets were found).
• 😀 The final output of the algorithm is a set of strong association rules that meet both the minimum support and confidence criteria, which can be used for further analysis or decision-making.

Q & A

What is the purpose of the Apriori algorithm in this video?

The Apriori algorithm is used to generate strong association rules from a given dataset. In this video, the algorithm helps discover frequent itemsets and then creates association rules based on the minimum support and confidence thresholds.

How are frequent itemsets generated in the Apriori algorithm?

Frequent itemsets are generated by first calculating the support count for individual items. Then, the minimum support is applied to filter out itemsets that do not meet the threshold. After that, combinations of frequent items are tested for higher-order frequent itemsets, and the process continues until no more frequent itemsets can be found.

What is the minimum support used in this example, and how is it calculated?

The minimum support given in this example is 40%. To calculate the minimum support count, the minimum support percentage is multiplied by the total number of instances in the dataset. For this dataset of 5 instances, the minimum support count is 2 (40% of 5).

Why is the confidence of an association rule important?

The confidence of an association rule represents the likelihood that an item in the rule will be found in a transaction if the other items are present. A higher confidence indicates a stronger association, and rules with a confidence greater than or equal to the minimum threshold are considered strong.

What is the formula for calculating the confidence of an association rule?

The confidence of an association rule is calculated as the frequency of both items (X and Y) occurring together, divided by the frequency of item X occurring alone. If the resulting value meets or exceeds the minimum confidence threshold, the rule is considered strong.

What is the difference between 'beach ocean' and 'ocean beach' in terms of association rule calculation?

'Beach ocean' and 'ocean beach' represent the same pair of items, but the order in which they appear can impact the calculation of confidence. For example, the rule 'beach ocean' may have a confidence of 50%, while 'ocean beach' could have a confidence of 100%, making it a stronger rule.

How is the process of generating two-item frequent itemsets described?

To generate two-item frequent itemsets, the distinct one-item frequent itemsets are paired up to form combinations. The support count for each combination is then calculated, and only those pairs with support counts equal to or greater than the minimum support are considered frequent.

Why can't a four-item frequent itemset be generated in this example?

A four-item frequent itemset cannot be generated because the support counts of the three-item sets do not meet the minimum support threshold. Once the maximum possible frequent itemset size is reached (in this case, two-item frequent itemsets), the process stops.

What does a confidence of 100% indicate for an association rule?

A confidence of 100% indicates that the association rule is very strong, meaning that if one item in the rule is present, the other item will always be present as well. This rule is considered very reliable.

How do you determine which association rules are 'strong' in the given example?

Association rules are considered strong if their confidence meets or exceeds the minimum threshold (70% in this example). Rules such as 'beach ocean' or 'ocean beach' with a confidence of 100% are strong, while others like 'sunshine holiday' with 50% confidence are not considered strong.