MS DEPARTMENT OF INFORMATION SYSTEM

Permanent URI for this community

https://escholar.umt.edu.pk/handle/123456789/7666

Browse

Browsing MS DEPARTMENT OF INFORMATION SYSTEM by Author "Faizan Irshad"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Use of AutoEndcoder & LSTM AutoEncoder for Real-time Anomaly Detection in Banking Transactio
    (UMT, Lahore, 2023) Faizan Irshad
    Anomaly is something that deviates from what is considered normal or expected. It can refer to a deviation from a rule or pattern, a strange or unusual occurrence, or a discrepancy in data. Anomaly detection in banking transactions involves identifying unusual or suspicious patterns in financial data including fraudulent or illegal activities, as well as potential errors or inaccuracies in the data. Banks process billions of transactions on daily basis worldwide and its practically impossible for the Banks to detect anomalies in the transactions using traditional rule based methods in real time environment. Artificial Intelligence offers a solution to this problem. We have introduced a novel approach using AutoEncoder and Long Short-Term Memory (LSTM) AutoEncoder for the detection of anomalous transaction in real life banking transactions dataset. One unique aspect of our model is also its implementation in an unsupervised manner separately for every account, allowing model to adapt to individual account behaviors and identify transactions that deviate from the norm for that particular account. Results of the unsupervised model were tested using labeled data. Our research shows that AutoEncoder outperformed LSTM AutoEncoder in terms of anomaly detection. The implementation of the proposed model on real life labelled transactions dataset has proved the effectiveness of proposed model in terms of real-time detection of anomalies in banking transactions which may be indicative of Fraud, Money laundering detection or Errors & Omissions.