Back to Portfolio

---

Predicting the next items in a grocery ordering & delivering app: Instacart Market Basket Analysis

Summary

Given Instacart’s open source dataset, 3 million Instacart Orders, in which there is information of millions of sales transactions that have been made through the app, it is of interest to try to determine if there are certain hidden patterns regarding consumer decisiones within the data.

This is how the data looks:

	order_id	product_id	add_to_cart_order	product_name	aisle_id	department_id
0	2	33120	1	Organic Egg Whites	86	16
1	26	33120	5	Organic Egg Whites	86	16
2	120	33120	13	Organic Egg Whites	86	16
3	327	33120	5	Organic Egg Whites	86	16
4	390	33120	28	Organic Egg Whites	86	16

The project focused on association rules for departments. The objective was to deliver a set of rules such as if the client just added a product from the Dairy department, it’s likely the next product will be from Bakery, for example. Making these suggestions would improve customer experience and increase sales.

This is how the recommendations would look:

	lift	antecedents_name	consequents_name
163	1.601961	[dairy eggs, deli]	[snacks, produce]
165	1.569408	[deli, produce]	[dairy eggs, snacks]
107	1.490181	[deli]	[snacks, produce]
145	1.486045	[dairy eggs, bakery]	[snacks, produce]
27	1.476974	[canned goods]	[pantry]
122	1.418792	[dairy eggs, bakery, produce]	[frozen]
135	1.409875	[dairy eggs, snacks, produce]	[frozen]
58	1.407305	[dairy eggs, deli]	[frozen]
161	1.402527	[dairy eggs, deli, produce]	[snacks]

References

• Data
• Apriori Algorithm
• Full Project

---

Back to Portfolio