"""

Creator: khanh.brandy

Created on 2020-06-30

"""

This source code works well with Neo4j 4.2 and is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY.

V1.1 DESCRIPTION

• User-Product Bipartite graph
• Use TRANSACTION (consider every single transaction and its amount)
• Do not use Neo4j procedure to calculate similarity score
• Cannot be used to extract and return ordered recommended items on a realtime basis

V1.2 UPDATES

• Inherit Bipartite graph structure from v1
• Try using TOTAL_AMT as the only relationship property with TOTAL_AMT = SUM(all transactions from node N to product P)
• Use Neo4j procedure to calculate similarity score
• Store similarity score between nodes as a new relationship [SIMILAR]
• Can be used to extract and return ordered recommended items on a realtime basis

V1.3 UPDATES

• Inherit graph structure, node properties (es SIMILAR) and relationship properties (es TOTAL_AMT) from v2
• Add more node properties then try using node embedded vector to calculate COSINE similarity score (from projected user-product Bipartite graph) along with (or instead of) using item-based vectors

V1.4 UPDATES

• Load original transactional graph then compute and store node embedded vector as a node property
• Build user-product Bipartite graph (using TOTAL_AMT as relationship)