Advanced rules
Market Basket Analysis in Python
Isaiah Hull
Visiting Associate Professor of Finance, BI Norwegian Business School
Overview of market basket analysis
- Standard procedure for market basket analysis.
- Generate large set of rules.
- Filter rules using metrics.
- Apply intuition and common sense.
Generating rules
- Number of rules grows exponentially in number of items.
- Most rules are not useful.
- Must apply initial round of filtering.
- Covered in chapter 3 using Apriori algorithm
How does filtering work?
| ID | antecedents | consequents | support | Lift |
|---|---|---|---|---|
| 1 | Harry Potter | The Hunger Games | 0.001 | 1.01 |
| 2 | Hunger Games | Twilight | 0.020 | 1.23 |
| 3 | Pride and Prejudice | The Hobbit | 0.030 | 1.05 |
| 4 | The Hobbit | Twilight | 0.015 | 0.85 |
| 5 | Harry Potter | The Hobbit | 0.010 | 1.07 |
Multi-metric filtering
| ID | antecedents | consequents | support | zhang |
|---|---|---|---|---|
| 1 | Harry Potter | The Hunger Games | 0.001 | -0.05 |
| 2 | Hunger Games | Twilight | 0.020 | 0.17 |
| 3 | Pride and Prejudice | The Hobbit | 0.030 | 0.06 |
| 4 | The Hobbit | Twilight | 0.015 | -0.04 |
| 5 | Harry Potter | The Hobbit | 0.010 | 0.34 |
Performing multi-metric filtering
print(rules.head())
antecedents consequents antecedent support ... support confidence ... conviction
0 Potter Hunger 0.48 ... 0.12 ... 0.26 ... 0.92
1 Potter Hunger 0.32 ... 0.12 ... 0.26 ... 0.92
2 Twilight Hunger 0.26 ... 0.09 ... 0.35 ... 1.04
3 Hunger Twilight 0.32 ... 0.09 ... 0.28 ... 1.03
4 Mockingbird Hunger 0.48 ... 0.10 ... 0.20 ... 0.85
Performing multi-metric filtering
# Select subset of rules with low consequent support.
rules = rules[rules['consequent support'] < 0.05]
print(len(rules))
12
# Select subset of rules with lift > 1.5.
rules = rules[rules['lift'] > 1.5]
print(len(rules))
2
Let's practice!
Market Basket Analysis in Python
