Item recommendation at brick-and-mortar convenience store

Problem description

The convenience store is divided into shop and restaurant part. There are three POS at the store. Two POS are located in the shop and one is located in the restaurant.
We want to recommend items for customers based on the past transactions and various features.

Demo1

Basic recommendation
Recommendation is performed with a model built over transaction data of all the POS. Model recommends items that are most-likely to be purchased according to shopping cart content and features. There are three main features considered:
1. season
2. weekday
3. part of a day
Demo2

Taxonomy-based recommendation
Recommendation is performed at two levels; i.e., item category level and item level. The former is more general and recommends item categories such as FOOD, DRINK, BEVERAGES. The latter recommends items within selected category. There are two recommendation models for each of the levels. Both are built over transaction data from all the three POS.
Demo3

Taxonomy-based recommendation per POS
This demo is similar to the Demo2 except that there are separate recommendation models for each POS.

Item recommendation at brick-and-mortar convenience store

Problem description

Demo1

Basic recommendation

Demo2

Taxonomy-based recommendation

Demo3

Taxonomy-based recommendation per POS