Publicação
Unsupervised anomaly detection of retail stores using predictive analysis library on SAP HANa XS advanced
| Resumo: | The retail industry is quite exposed to fraudulent situations. Daily, thousands of transactions are processed, which may include some frauds difficult to detect, mainly when the perpetrators are the own employees at the retail stores. Large retailers with several stores across different locations may have considerable difficulty in detecting frauds involving their cashiers since they have to take into account different contexts of operation. To reduce fraud losses, retailers get an overview of the transactions in each store to filter the ones that look suspicious deviating from what would be normal. Data mining algorithms can be useful to detect anomalies, differentiating the normal from the abnormal. This study adopted the k-Means clustering algorithm for anomaly detection on a sample of 90 stores in a large food retail chain, revealing the existence of some outliers in the data. The anomaly detection process was fully implemented in SAP HANA XS Advanced using the Predictive Analysis Library (PAL). In the end, it was possible to identify the stores with abnormal behavior and conclude for the usefulness and ease of use of such a library, despite some lack of documentation to use it. |
|---|---|
| Autores principais: | Oliveira, João Pedro |
| Outros Autores: | Sousa, Rui Dinis |
| Assunto: | K-means clustering PAL Retail SAP HANA XSA Unsupervised anomaly detection |
| Ano: | 2021 |
| País: | Portugal |
| Tipo de documento: | comunicação em conferência |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | The retail industry is quite exposed to fraudulent situations. Daily, thousands of transactions are processed, which may include some frauds difficult to detect, mainly when the perpetrators are the own employees at the retail stores. Large retailers with several stores across different locations may have considerable difficulty in detecting frauds involving their cashiers since they have to take into account different contexts of operation. To reduce fraud losses, retailers get an overview of the transactions in each store to filter the ones that look suspicious deviating from what would be normal. Data mining algorithms can be useful to detect anomalies, differentiating the normal from the abnormal. This study adopted the k-Means clustering algorithm for anomaly detection on a sample of 90 stores in a large food retail chain, revealing the existence of some outliers in the data. The anomaly detection process was fully implemented in SAP HANA XS Advanced using the Predictive Analysis Library (PAL). In the end, it was possible to identify the stores with abnormal behavior and conclude for the usefulness and ease of use of such a library, despite some lack of documentation to use it. |
|---|