Market Basket Analysis

Market basket analysis with Python. Generate synthetic retail transactions, mine frequent itemsets using Apriori or FP‑Growth, derive association rules, and produce clean outputs and charts.

---

Features

• Synthetic transactions generator with realistic co‑purchases (e.g., Laptop → Mouse/Keyboard/Bag)
• Frequent itemsets via Apriori or FP‑Growth (mlxtend)
• Association rules with support, confidence, lift, leverage, conviction
• Reproducible artifacts: CSVs for item supports, frequent itemsets, association rules
• Visual: Top‑N items by support (Matplotlib)
• All outputs saved under outputs/

---

Project Structure

market-basket-analysis/
├─ README.md
├─ LICENSE
├─ requirements.txt
├─ data/
│  └─ generate_transactions.py
├─ src/
│  ├─ market_basket.py
│  └─ utils.py
└─ outputs/
   └─ figures & reports (auto-created)

---

Setup

python -m venv .venv
# Windows:
.venv\Scripts\activate
# macOS/Linux:
source .venv/bin/activate 
pip install -r requirements.txt

---

Generate Synthetic Data

python data/generate_transactions.py   --start 2024-01-01 --end 2024-12-31   --customers 800 --avg_per_day 120 --seed 42   --out data/transactions.csv

---

Run Market Basket Analysis

Apriori:

python src/market_basket.py   --input data/transactions.csv --outdir outputs   --algo apriori --min_support 0.02 --metric lift --min_threshold 1.1 --top_n 12

FP‑Growth:

python src/market_basket.py   --input data/transactions.csv --outdir outputs   --algo fpgrowth --min_support 0.02 --metric lift --min_threshold 1.1 --top_n 12

Outputs

• outputs/item_support.csv
• outputs/frequent_itemsets_<algo>.csv
• outputs/association_rules_<algo>.csv
• outputs/fig_top_items.png
• outputs/summary.json

---

---

Example Results (from a real run)

Top rule (by lift)

Antecedents	Consequents	Support	Confidence	Lift	Conviction
Keyboard, Laptop Bag, USB-C Hub	Laptop, Mouse	0.024	0.684	5.095	2.743

Interpretation: baskets containing Keyboard + Laptop Bag + USB‑C Hub are ~5× more likely than random to also include Laptop + Mouse.

Top‑N item supports

Top rule (by lift)

Antecedents	Consequent	Support	Confidence	Lift
Laptop, Mouse	Keyboard	0.036	0.58	2.41

Interpretation: if a basket contains Laptop + Mouse, it’s ~2.4× more likely to also contain Keyboard than at random.

Top‑N item supports (chart saved to outputs/fig_top_items.png).

---

Methods

• Frequent Itemset Mining: Identify groups of products that co‑occur in transactions. Two algorithms are provided:
  • Apriori: generates candidate itemsets level‑wise using the downward‑closure property, pruning infrequent candidates early.
  • FP‑Growth: compresses the dataset into an FP‑tree to mine itemsets without explicit candidate generation. Efficient on dense data.
• Association Rules: For each frequent itemset, derive rules X → Y (with X ∩ Y = ∅) and compute:
  • Support: P(X ∪ Y)
  • Confidence: P(Y|X)
  • Lift: P(Y|X) / P(Y) ( >1 indicates positive association )
  • Leverage and Conviction for additional signal.

---

Notes

• The synthetic generator biases realistic attachments (e.g., Laptop → Mouse/Keyboard/Bag) so mined rules are interpretable.
• Tune --min_support and --min_threshold for sparser vs. denser rule sets.
• For large datasets, FP‑Growth is typically faster than Apriori.

---

About

Market basket analysis with Python. Generate synthetic retail transactions, mine frequent itemsets using Apriori/FP‑Growth, derive association rules, and export figures and CSVs for quick portfolio demos.