Anomaly Detection in Sales Transactions for FMCG (Fast Moving Consumer Goods) Distribution
Abstract
In today’s era, companies operating in the FMCG industry played an important role in society, especially regarding the distribution of goods used in daily life, which were distributed directly from factories or principals. Despite rapid technological advancements, many distribution companies in Indonesia still relied on human labor and conducted distribution processes manually. Concerns about inaccuracies in employee actions and other detrimental activities such as embezzlement, fraud, and so on, drove companies to undertake digital transformation processes. To reduce these risks, some FMCG companies had already implemented systems to monitor distribution activities and customer payment processes. However, another issue arose due to the limited number of employees available to conduct professional audits, resulting in suboptimal monitoring processes and increased risks of integrity issues or fraud committed by employees. To address this, the implementation of an Autoencoder system was utilized to help companies detect fraudulent activities, particularly in the sales domain. Referring to this study, it showed that the implementation of machine learning technology, such as Autoencoders, yielded positive results and was considered effective in detecting suspicious activities, especially in large transaction datasets. The Autoencoder system utilized in this research was developed using TensorFlow, showing promising results in detecting fraudulent transactions in the company. Additionally, the model was able to train on 80% of the data and was tested on the remaining 20%. According to the outcome, approximately 6.664% of transactions were predicted to be fraudulent. Based on the results, this research showed that the implementation of the AutoEncoder system had proven to be effective in helping the organization prevent and protect against potential non-compliant activities. This proof could be used as a learning opportunity for other organizations facing similar challenges.
Article Metrics
Abstract: 40 Viewers PDF: 16 ViewersKeywords
Full Text:
PDFRefbacks
- There are currently no refbacks.
Journal of Applied Data Sciences
ISSN | : | 2723-6471 (Online) |
Organized by | : | Computer Science and Systems Information Technology, King Abdulaziz University, Kingdom of Saudi Arabia. |
Website | : | http://bright-journal.org/JADS |
: | taqwa@amikompurwokerto.ac.id (principal contact) | |
support@bright-journal.org (technical issues) |
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0