Geavanceerd filteren met meerdere metrieken

Eerder gebruikten we data van een online cadeauwinkel met gadgets om antecedenten te vinden die we konden gebruiken om een gerichte consequent te promoten. Omdat de set met potentiële regels groot was, moesten we vertrouwen op het Apriori-algoritme en filteren met meerdere metrieken om die te verkleinen. In deze oefening bekijken we de volledige set regels en zoeken we een bruikbare regel, in plaats van een specifieke antecedent te targeten.

Let op: de data is geladen, voorbewerkt en one-hot gecodeerd, en beschikbaar als onehot. Daarnaast zijn apriori() en association_rules() geïmporteerd uit mlxtend. In deze oefening pas je het Apriori-algoritme toe om frequente itemsets te identificeren. Daarna haal je de set met associatieregels uit de itemsets en pas je filteren met meerdere metrieken toe.

Deze oefening maakt deel uit van de cursus

Market Basket Analysis in Python

Cursus bekijken

Oefeninstructies

Pas het Apriori-algoritme toe op de one-hot gecodeerde itemsets met een minimale supportdrempel van 0,001.
Haal associatieregels op met een minimale supportdrempel van 0,001.
Stel antecedent_support in op 0,002 en consequent_support op 0,01.
Stel confidence hoger in dan 0,60 en lift hoger dan 2,50.

Praktische interactieve oefening

Probeer deze oefening eens door deze voorbeeldcode in te vullen.

# Apply the Apriori algorithm with a minimum support threshold of 0.001
frequent_itemsets = ____(onehot, min_support = ____, use_colnames = True)

# Recover association rules using a minium support threshold of 0.001
rules = ____(frequent_itemsets, metric = '____', min_threshold = 0.001)

# Apply a 0.002 antecedent support threshold, 0.60 confidence threshold, and 2.50 lift threshold
filtered_rules = rules[(rules['antecedent support'] > ____) &
						(____['consequent support'] > 0.01) &
						(rules['____'] > ____) &
						(____ > 2.50)]

# Print remaining rule
print(filtered_rules[['antecedents','consequents']])

Code bewerken en uitvoeren

Deze oefening maakt deel uit van de cursus

Market Basket Analysis in Python

SkillTag.level.intermediateSkillTag.label

4.9+

Begin de cursus gratis

In this chapter, you’ll learn the basics of Market Basket Analysis: association rules, metrics, and pruning. You’ll then apply these concepts to help a small grocery store improve its promotional and product placement efforts.

Exercise 1: What is market basket analysis?Exercise 2: The basics of market basket analysis Exercise 3: Cross-selling products Exercise 4: Identifying association rules Exercise 5: Multiple antecedents and consequents Exercise 6: Preparing data for market basket analysis Exercise 7: Generating association rules Exercise 8: The simplest metric Exercise 9: One-hot encoding transaction data Exercise 10: Computing the support metric

Association rules tell us that two or more items are related. Metrics allow us to quantify the usefulness of those relationships. In this chapter, you’ll apply six metrics to evaluate association rules: supply, confidence, lift, conviction, leverage, and Zhang's metric. You’ll then use association rules and metrics to assist a library and an e-book seller.

Exercise 1: Confidence and lift Exercise 2: Recommending books with support Exercise 3: Refining support with confidence Exercise 4: Further refinement with lift Exercise 5: Leverage and conviction Exercise 6: Lift versus leverage Exercise 7: Computing conviction Exercise 8: Computing conviction with a function Exercise 9: Promoting ebooks with conviction Exercise 10: Association and dissociation Exercise 11: Computing association and dissociation Exercise 12: Defining Zhang's metric Exercise 13: Applying Zhang's metric Exercise 14: Advanced rules Exercise 15: Filtering with support and conviction Exercise 16: Using multi-metric filtering to cross-promote books

The fundamental problem of Market Basket Analysis is determining how to translate vast amounts of customer decisions into a small number of useful rules. This process typically starts with the application of the Apriori algorithm and involves the use of additional strategies, such as pruning and aggregation. In this chapter, you’ll learn how to use these methods and will ultimately apply them in exercises where you assist a retailer in selecting a physical store layout and performing product cross-promotions.

Exercise 1: Aggregatie Exercise 2: Aggregatie uitvoeren Exercise 3: Een aggregatiefunctie definiëren Exercise 4: Het Apriori-algoritme Exercise 5: Pruning en Apriori Exercise 6: Frequent itemsets identificeren met Apriori Exercise 7: Een supportdrempel selecteren Exercise 8: Basisresultaten van Apriori snoeien Exercise 9: Associatieregels genereren Exercise 10: Prunen met lift Exercise 11: Prunen met confidence Exercise 12: Geavanceerd snoeien van Apriori-resultaten Exercise 13: Aggregatie en filteren Exercise 14: Zhangs regel toepassen Exercise 15: Geavanceerd filteren met meerdere metrieken

Huidige oefening

In this final chapter, you’ll learn how visualizations are used to guide the pruning process and summarize final results, which will typically take the form of itemsets or rules. You’ll master the three most useful visualizations -- heatmaps, scatterplots, and parallel coordinates plots – and will apply them to assist a movie streaming service.

Exercise 1: Heatmaps Exercise 2: Visualizing itemset support Exercise 3: Heatmaps with lift Exercise 4: Interpreting heatmaps Exercise 5: Scatterplots Exercise 6: Pruning with scatterplots Exercise 7: Optimality of the support-confidence border Exercise 8: Parallel coordinates plot Exercise 9: Using parallel coordinates to visualize rules Exercise 10: Refining a parallel coordinates plot Exercise 11: Congratulations!