A carregar...
Publicação
Improving merchant identification in transaction data using Word2Vec and GloVe embeddings
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.65 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Merchant identification and aggregation in financial transaction data is crucial for ensuring
fraud detection and seamless payment processing. Inconsistent merchant data often results in
legitimate transactions being flagged as fraudulent or fraudulent activities going undetected.
This thesis explores the use of NLP techniques, including Word2Vec and GloVe, to enhance
merchant identification by clustering similar entities. A filtering process for dominant mer chants was implemented, addressing clustering limitations. While Word2Vec showed potential
for capturing contextual nuances, neither method outperformed an existing string-matching
model. Future research should explore advanced NLP models, larger datasets, and address the
scarcity of public data for merchant aggregation.
Descrição
Palavras-chave
NLP Word embeddings Transaction data Merchant identification Word2Vec GloVe