The objective of the project is to identify instances of circular trading in a given set of transactions. A dataset containing 32449 unique buyers and sellers containing a total of 102256 transactions was used.
For analysis of transactions, we construct a “trade-network” in the form of a directed graph using the provided dataset. Each trader who bought/sold commodities enters the graph network as a node. For each transaction in the dataset, we construct a directed edge originating from the buyer’s node to the seller’s node. A weight is also attributed with each edge in the graph which is equal to the amount of money involved in the corresponding transaction.
Our approach is based on using heuristics from existing graph clustering algorithms to identify potentially fraudulent clusters.
Some examples of potential candidates for circular trade networks identified by our algorithm are given below. The number on the nodes indicates the buyer/seller ID.
Ninad Akolekar
Faizul Haq
Ayush Pateria
Dr. Sobhan Babu
Associate Professor
Department of Computer Science & Engineering
IIT Hyderabad
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent