Product Name Match and Item Relationship Exploration on Transaction Logs

Repository for the MS Analytics Practicum Project for CSE/ISYE/MGMT 6748 Fall '20. This project was completed along with Hannah Hamilton, Qihang Zhang, Jingyu Li and Xinze Wang.

1. Data

• All dataset should be put in "/jupyter_notebooks/raw_data/".
• Dataset "/jupyter_notebooks/raw_data/third_party/food.csv": grocery product list by USDA (https://www.usda.gov/content/usda-open-data-catalog)
• Dataset "/jupyter_notebooks/raw_data/third_party/generic_food.csv": grocery product list by Instacart.
• Dataset "/jupyter_notebooks/raw_data/third_party/instacart/*": more product information datasets by Instacart.
• Dataset "/jupyter_notebooks/raw_data/third_party/grocery.com/*": files from the Open Grocery Database Project by grocery.com.
• Model Outputs "/jupyter_notebooks/raw_data/model_outputs/*": this folder contains some files output by our models.

2. Problem 1 setup: Product name match

• Step1: run notebook "/jupyter_notebooks/problem1_name_match/master_catalog_generation.ipynb"
• Step2: run notebook "/jupyter_notebooks/problem1_name_match/similarity_based_approach.ipynb.ipynb"

3. Problem 2 setup: Item recommendation

• Apriori algorithm for complementary products: run notebook "/jupyter_notebooks/problem2_product_recommendation/apriori_optimized_min_support_of_top_10%_cleaned_v2.ipynb" and "/jupyter_notebooks/problem2_product_recommendation/apriori_optimized_small_transactions_min_support_of_top_10%_cleaned_v2.ipynb".
• Transaction context co-occurrence for substitute products: run notebook "/jupyter_notebooks/problem2_product_recommendation/substitute_exploration_transaction_context.ipynb".
• Graph-based GCN for substitute products: run notebook "/jupyter_notebooks/problem2_name_match/graph_based_item_recommendation.ipynb"