Fraud Detection using Risk Based Binary Classification

In the era of complicated information, consumer behavior mining is of great significance to enterprises and markets, however, the current information processing technology is difficult to comprehensively mine and segment consumer data. The study aims to conduct an effective analysis of consumer behavior. And for the shortcomings of the current consumer behavior mining algorithms, the study proposes an improved consumer behavior data mining algorithm based on the map reduce model. After the experimental analysis, the results revealed that the research algorithm has the closest Mahalanobis Distances compared to the two algorithms, fuzzy C-means and density-based spatial clustering of application with noise, indicating that the research algorithm is more effective in clustering. The average clustering accuracy of K-means clustering algorithm (K-means) based on Andersori’s Iris data seto dataset was 93.2%, and the average clustering accuracy of the two datasets Glass and Wine was 94.3% and 93.8%, respectively. The research methodology categorized consumers into three classes based on their transaction frequency and transaction amount. Among the consumers in cluster 1, the total transaction amount was in the range of 0.62–0.82, the transaction frequency was between 0.41 and 0.72, and the number of transactions was between 0.72 and 0.94, which shows that the consumers in this cluster belonged to the group of moderately active and high consumption. The above data indicate that this method, through the collaborative optimization of MapReduce and K-means, has an accuracy rate of over 90% in cross-industry scenarios. It effectively solves the problems of low efficiency, poor accuracy and weak adaptability of traditional algorithms, providing a quantifiable technical solution for the research of consumer behavior.

Smart contracts are the cornerstone of decentralized applications and financial protocols, extending digital currency transactions but also introducing serious security challenges that cause substantial economic losses. Existing solutions primarily target code‐level vulnerabilities, which constitute only a portion of all security incidents. Though prior research conducts static analysis across contract lifecycles, they lack the characteristic analysis of vulnerabilities in each stage and the distinction between the vulnerabilities. This paper presents the first empirical study on the security of smart contracts throughout their lifecycle, including deployment and execution, upgrade, and destruction stages. We propose two feature categories: transaction features (contract lifespan and four dynamic features: transaction numbers, transaction amounts, neighbour numbers, and the number of old and new neighbours) and ego network properties (temporal network density and local clustering coefficient). These seven features capture behavioural patterns across dimensions, including activity duration, transaction frequency, financial liquidity, interaction breadth, dynamic interaction changes, and structural properties of the transaction network. Finally, five machine learning models—logistic regression, random forest, support vector machine, decision tree, and K‐nearest neighbours—are used to identify vulnerabilities at different stages. Results show that vulnerable contracts exhibit distinct transaction features and ego network properties at different lifecycle stages.

Bank fraud is an increasingly prevalent issue, causing substantial financial losses for both financial institutions and customers. Also, with the increasing prevalence of bank fraud, especially facilitated by online banking and digital payments, there is a pressing need for potential expertise deployment and advanced fraud detection techniques in real-world banking systems. This study specified the Bayesian Network models for detecting fraudulent bank transactions. Bayesian Networks offer a probabilistic graphical modeling approach that can effectively capture complex relationships and dependencies within financial data. Thus, the research aimed at developing a custom Bayesian network model trained on a large dataset of bank transactions comprising over one million bank transactions to classify instances as fraudulent or non-fraudulent. Python 3.11.10 was used for the graphical representations. Down-sampling was employed to reduce the dataset from its initial size of 1 million observations to 17,650 observations, ensuring a balanced representation of both fraud and non-fraud instances. Various estimation techniques: Maximum Likelihood, Bayesian Networks, and Expectation Maximization were employed and evaluated to learn the model parameters. The model's performance was measured using metrics: accuracy, precision, recall, F1-score, and ROC-AUC. The Bayesian Networks estimator achieved an overall accuracy of 66.18%, precision of 67.73%, F1-score of 64.06%, ROC-AUC of 66.13%, Recall of 60.77%, Sensitivity of 60.77% and Specificity of 71.50%. Also, the Maximum Likelihood achieved an overall Accuracy of 66.77%, Precision of 69.22%, F1-score of 63.96%, ROC-AUC of 66.71%, Recall of 59.45%, Sensitivity of 59.45% and Specificity of 73.97%. Likewise, the Expectation Maximization achieved an overall Accuracy of 66.83%, Precision of 69.31%, F1-score of 64.00%, ROC-AUC of 66.77%, Recall of 59.45%, Sensitivity of 59.45% and Specificity of 74.09%. On the confusion matrix, the model correctly classified 1272 instances as Non-Fraudulent transactions (True Negative). Also, it was observed that the model incorrectly classified 507 instances as fraudulent transactions when they were Non-Fraudulent (False Positive). The model similarly incorrectly classified 687 instances as Non-Fraudulent transactions when they were Fraudulent (False Negative). Finally, it correctly classified 1064 instances as fraudulent transactions (True Positive). These results demonstrated the Bayesian Network's ability to identify fraudulent transactions while minimizing false alarms accurately. The findings highlight the potential of Bayesian Networks as a robust framework for fraud detection in the banking sector, contributing to enhanced security and reduced financial losses. The developed model can be introduced into existing banking systems to strengthen fraud prevention strategies.

Fraudulent transactions and the methods to detect them are an an important issue for financial organizations globally. The requirement for progressive fraud detection systems to protect properties and maintain customer trust is predominant for financial organizations, but particular factors make the development of efficient and effective fraud detection models a challenge. Deep Learning (DL) has greatly improved fraud detection accuracy by detecting intrinsic patterns, whereas interpretability techniques improve transparency and build trust by making predictions understandable to experts. This study presents a Fraud Detection System using Recursive Feature Elimination and Waterwheel Plant Optimization (FDS-RFEWPO) model for financial transactions. The aim is to perform a comprehensive evaluation of fraud detection in high-dimensional financial transactions using advanced techniques. Initially, the FDS-RFEWPO technique follows min-max-based data pre-processing to normalize the input data. For the feature selection process, the FDS-RFEWPO model employs the Recursive Feature Elimination (RFE) technique to select the most relevant features from the dataset. Furthermore, the Variational Autoencoder/Wasserstein Autoencoder (VAE/WAE) model is employed for fraud detection and classification. To further enhance model performance, the Waterwheel Plant Optimization (WPO) technique is employed for hyperparameter tuning, ensuring the selection of optimal parameters that contribute to improved accuracy. Finally, the Explainable Artificial Intelligence (XAI) technique applies Local Interpretable Model-Agnostic Explanations (LIME) to improve the transparency, interpretability, and trustworthiness of Artificial Intelligence (AI) methods by making their decision-making procedures clear to humans. To evaluate the performance of the FDS-RFEWPO model, a comprehensive experimental analysis is conducted using a financial fraud detection dataset. The comparison study of the FDS-RFEWPO model demonstrates a superior accuracy of 97.41% over existing techniques.

In this digital world, numerous credit card-based transactions take place all over the world. Concomitantly, gaps in process flows and technology result in many fraudulent transactions. Owing to the spurt in the number of reported fraudulent transactions, customers and credit card service providers incur significant financial and reputation losses respectively. Therefore, building a powerful fraud detection system is paramount. It is noteworthy that fraud detection datasets, by nature, are highly unbalanced. Consequently, almost all of the supervised classifiers, when built on the unbalanced datasets, yield high false negative rates. But, the extant oversampling methods while reducing the false negatives, increase the false positives. In this paper, we propose a novel data oversampling method using Generative Adversarial Network (GAN). We use GAN and its variant to generate synthetic data of fraudulent transactions. To evaluate the effectiveness of the proposed method, we employ machine learning classifiers on the data balanced by GAN. Our proposed GAN-based oversampling method simultaneously achieved high precision, F1-score and dramatic reduction in the count of false positives compared to the state-of-the-art synthetic data generation based oversampling methods such as Synthetic Minority Oversampling Technique (SMOTE), Adaptive Synthetic Sampling (ADASYN) and random oversampling. Moreover, an ablation study involving the oversampling based on the ensemble of SMOTE and GAN/WGAN generated datasets indicated that it is outperformed by the proposed methods in terms of F1 score and false positive count.

The escalating threat of fraud in financial institutions is a global issue, with the Malaysian sector being no exception. This study focuses on the implementation and efficacy of Data Mining and Machine Learning methodologies in identifying and mitigating fraudulent activities within these institutions. The paper critically reviews existing literature, bridging the gap between advanced technology application and fraud management. Fraudulent transactions in the financial sector are dynamic and sophisticated, requiring advanced detection techniques. Traditional approaches often struggle to manage this complexity effectively, demonstrating a need for more advanced and adaptive strategies. This is where Data Mining and Machine Learning techniques, renowned for their predictive and analytical prowess, can significantly contribute. Data Mining, the process of uncovering patterns and correlations within large datasets, is a useful tool for detecting anomalies that may suggest fraud. The study assesses various data mining techniques, such as clustering, classification, and association, and explores their application in detecting fraudulent transactions. Findings indicate that these techniques can substantially enhance fraud detection rates while minimizing false positives. Furthermore, Machine Learning, an artificial intelligence subset, has shown immense potential in fraud detection. Its ability to learn from and make decisions based on data makes it a viable solution for fraud detection. This paper explores both supervised and unsupervised learning algorithms and their efficacy in identifying fraud in the Malaysian financial sector. Results suggest that machine learning models, when correctly implemented, can significantly improve the accuracy of fraud detection. The review underscores the importance of employing advanced technologies like Data Mining and Machine Learning to combat financial fraud effectively. It also suggests future research directions, emphasizing the need for context-specific, localized models considering Malaysia's unique socio-economic environment. Moreover, the development of hybrid models, integrating both data mining and machine learning, could offer improved results. In conclusion, this study sets a precedent for further exploration into the application of advanced analytical tools in fraud detection in the Malaysian financial sector. The potential these technologies offer for improving accuracy and adaptability in fraud detection systems is substantial and warrants thorough investigation. Keywords: Fraud Detection, Malaysian Financial Institutions, Data Mining, Machine Learning, Financial Fraud Management

The expansion of electronic payments in the retail sector has been accompanied by an increase in advanced payment frauds. Traditional rule based legacy static fraud detection methods, in most cases, proved inadequate to address dynamically changing and constantly evolving fraudster practices. This paper discusses an efficient real time fraud detection system through a combination of Machine Learning (ML) algorithms and real time data stream processing technology. The study is helpful in the sense that it has a constructed synthetic transaction data set which was specially designed and built to be utilized for the very specific purpose of simulating real retail payment scenarios. The data have 489 samples with columns of transaction amount, merchant category, time of day, and frequency of transactions. We test and use a subset of the ML algorithms, like Random Forest and Gradient Boosting, to identify fraud effectively with very low latency. The solution takes consumption transaction as an Apache Kafka stream simulated, real time feature creation, and scoring of transactions using trained ML models. It depicts how ensemble methods, of which Gradient Boosting is a fair example, are more precise in recall and accuracy and minimize false negatives and false positives to a significant degree. This research acknowledges the prospect of a streaming paradigm and sophisticated ML in developing adaptive and real time fraud detection systems of high value addition to the security framework of modern retail payment systems.

Background: Fraud in financial transaction is at the root of corruption issues recorded in organization. Detecting fraud practices has become increasingly complex and challenging. As a result, auditors require precise analytical tools for fraud detection. Grouping financial transaction data using K-Means Clustering algorithm can enhance the efficiency of applying Benford Law for optimal fraud detection. Objective: This study aimed to introduce Multiple Benford Law Model for the analysis of data to show potential concealed fraud in the audited organization financial transaction. The data was categorized into low, medium, and high transaction values using K-Means Clustering algorithm. Subsequently, it was reanalyzed through Multiple Benford Law Model in a specialized fraud analysis tool. Methods: In this study, the experimental procedures of Multiple Benford Law Model designed for public sector organizations were applied. The analysis of suspected fraud generated by the toolkit was compared with the actual conditions reported in audit report. The financial transaction dataset was prepared and grouped into three distinct clusters using the Euclidean distance equation. Data in these clusters was analyzed using Benford Law, comparing the frequency of the first digit’s occurrence to the expected frequency based on Benford Law. Significant deviations exceeding ±5% were considered potential areas for further scrutiny in audit. Furthermore, the analysis were validated by cross-referencing the result with the findings presented in the authorized audit organization report. Results: Multiple Benford Law Model developed was incorporated into an audit toolkit to automated calculations based on Benford Law. Furthermore, the datasets were categorized using K-Means Clustering algorithm into three clusters representing low, medium, and high-value transaction data. Results from the application of Benford Law showed a 40.00% potential for fraud detection. However, when using Multiple Benford Law Model and dividing the data into three clusters, fraud detection accuracy increased to 93.33%. The comparative results in audit report indicated a 75.00% consistency with the actual events or facts discovered. Conclusion: The use of Multiple Benford Law Model in audit toolkit substantially improved the accuracy of detecting potential fraud in financial transaction. Validation through audit report showed the conformity between the identified fraud practices and the detected financial transaction. Keywords: Fraud Detection, Benford’s Law, K-Means Clustering, Audit Toolkit, Fraudulent Practices.

BI-RTGS is transferring funds with an essential function in the payment system and third-party funds as a factor to meet the current account balance which is a condition for BI-RTGS transactions. Bank financing is considered to be able to affect the number of BI-RTGS transactions. This research used 14 Islamic commercial banks registered as BI-RTGS participants. The results show that Islamic Commercial Bank Third Party Funds have a partial effect on the nominal amount of BI-RTGS transactions. Islamic Commercial Bank Financing has a partial effect on the nominal amount of BI-RTGS transactions. Third Party Funds and Islamic Commercial Bank Financing simultaneously affect the nominal amount of BI-RTGS transactions. In the determination coefficient test, the R Square value of third-party funds is 11.3% and financing is 4.1%. This means that the influence of third party funds is greater than the amount of the influence of financing on the nominal amount of BI-RTGS transactions in Islamic Banking in Indonesia.

Data imbalance is one of the main issues in online transactions fraud detection systems where the majority class (i.e., genuine transactions) dominates the minority class (i.e., fraud transactions). Deep learning-based fraud detection systems have the capability to learn efficiently with imbalanced data. Thresholding with Deep Learning is still understudied. Thresholding or threshold moving alters the decision threshold of the learning model to change its output rather than changing the data distribution or learning of the model. A fraud detection system is considered efficient if it can detect the maximum number of frauds and maintain the overall performance of model. Thus, this paper presents a comparison of three thresholding methods based on Receiver Operating Characteristic (ROC) Curve i.e., Closest to (0,1) criteria, Youden Index (J), max -G-Mean in deep-learning based online transactions fraud detection system. The experimental results show that closest to (0,1) criterion has achieved maximum fraud detection rate i.e. True Positive Rate (TPR) as compared to the other two methods. Hence, selecting the right decisionthresholding method helps in achieving better results.

The existence of fraud in credit card transactions causes many financial losses leading to customers’ loss of trust. Fraud detection methods based on machine learning techniques prevent such losses. Despite the literature on fraud detection, there is a lack of algorithms that detect fraud with acceptable performance in the credit card fraud detection field. Therefore, this study proposed a comprehensive ensemble-based method using an efficient weighted voting strategy for credit card fraud detection that can address the previous algorithms’ weaknesses. First, since the dataset is imbalanced, the proposed method balanced the dataset by stratifying it into three different proportions of normal and fraudulent transactions (1 to 1, 1 to 4 and 1 to 9 ratios). The features in each dataset are ranked by four feature-ranking methods, and the Random Forest classifier is applied to each of them for selecting the essential features. Afterward, using the seven base classifiers and the chosen features, 12 ensembles have been developed. Besides, a weighted voting strategy is proposed, and the fraudulent transactions are detected through voting based on the base classifiers’ and ensembles’ weights, which are calculated by their performance. The computational results indicated that the suggested Eclf10 is the best ensemble and its Logistic Regression classifier also has the best performance among other base classifiers. The Eclf10 leads to 99.97% accuracy, 87.78% precision, 97.70% recall, 92.21% F1-score and 95.634% F2-score, which has a superiority over the previous ensemble-based methods (e.g. majority voting ensemble, stacking classifier, Adaboost, Gradient Boosting).

As customers tendencies learn rapidly about online transactions due to the outbreak of the Covid-19 pandemic, fraudulent transactions that mimic legitimate transactions will continue to increase unabated. Over the years, fraudulent transactions have been committed but the supervised and unsupervised machine learning classifiers and fraud detection techniques could not stop the fraud from occurring at the onset. However, fraudulent cases are detected after post-occurrence thereby recognizing these techniques as non-smart detection models because fraud could only be detected using past data sets. Various fraud detection or classifier techniques using machine learning techniques such as decision trees, random forests, k-means, and neural networks have been studied. Also, the possibility of overfitting due to the branch growth associated with the decision trees is a classifier performance problem. The application of the probability axiomatic metric (PAM) to determine the accuracy of the models is a classification evaluation problem because of the relatively moderate error rate. To overcome the above shortfalls, we proposed a threshold Mahalanobis distance (TMD) classifier and adopted BETH as a performance evaluation metric. The result revealed that TMD is comparable to neural network (NN), k-means, isolation forest, XGBoost, and random forest based on the credit card data set. The analysis showed that TMD and NN have higher efficiency compared to other classifiers. This study demonstrated that the TMD classifier is unique and could be applied to extract legitimate and illegitimate transactions from customer transaction data sets. The study concludes that TMD is robust and comparable to NN and the classification errors associated with BETH are minimal compared to PAM for both TMD and NN.

As dependency on credit cards and online transactions is increasing, the scope for unauthorized payments or fraudulent transactions is on the rise. Huge financial losses are borne by banks every year due to fraudulent transactions. The process of identifying such transactions among millions of transaction records can be perfected using machine learning models, even more so by ensembling them. This study takes a real-time financial dataset and analyzes the efficiency of various machine learning models in predicting fraudulent transactions. Individual ML models are also analysed, ensembled and tested on the dataset. Upon extensive analysis and research, it is concluded that an ensembled model of LightGBM, decision tree classifier and XGBoost gave the highest AuC score of 87%. Further, a user-interactive tool is also created, which predicts fraudulent transactions from real-time transactions. The tool makes use of the model with highest prediction efficiency and presents it in an interactive manner. This study and practical demonstration show advancement in prediction of fraudulent transactions when compared to existing ML models and its user-friendly software incorporation.KeywordsFraud detectionEnsembleAccuracyFraudulent transactionBankingMachine learning model

Detecting online fraudulent trading in the realm of Fintech presents several challenges, primarily due to the dynamic nature of financial markets and the evolving tactics of fraudsters. Traditional machine learning algorithms trained on unbalanced datasets tend to bias towards the majority class (legitimate transactions) and may overlook minority class (fraudulent transactions) patterns. This bias can lead to poor performance in detecting fraudulent activities. The choice of sampling technique (e.g., oversampling, undersampling, SMOTE) can significantly impact model performance. However, selecting the appropriate sampling strategy requires domain knowledge and experimentation, which can be time-consuming and resource-intensive. This work utilizes Artificial Bee Colony (ABC)-based sampling to tackle class imbalance in credit card fraud detection. By generating realistic synthetic fraud samples, ABC-sampling helps the model learn fraudulent patterns more effectively without favoring non-fraudulent transactions. Inspired by the foraging behavior of bees, the process involves exploring existing fraud patterns, selecting the most relevant ones, creating synthetic fraud samples, and refining them to ensure they closely resemble real fraud cases while preserving key features that distinguish fraud from regular transactions. This method enhances fraud detection accuracy and minimizes false alarms, making the system more reliable. The suggested model employs anomaly detection algorithm to identify unusual or fraudulent trading activities in which it creates behavioral profiles for individual traders based on their historical trading activities and utilizes machine learning algorithm to cluster traders into groups based on similar behavior patterns. Then it identifies characteristic features of fraudulent traders such as sudden changes in trading volume, irregular trading hours, or trading activities inconsistent with their profile. By analyzing patterns and anomalies in traders’ behavior, these approaches can effectively identify suspicious activities indicative of fraudulent behavior. Extensive performance studies demonstrate that the proposed algorithm significantly outperforms the state-of-the-art methods by 10% in terms of accuracy.

Credit card fraud is a common occurrence in today’s society because the majority of us use credit cards as a form of payment more frequently. This is the outcome of developments in technology and an increase in online transactions, which have given rise to frauds that have caused significant financial losses. In order to detect fraud in credit card transactions, efficient and effective approaches are needed. In this study, we developed a hybrid CNN‐SVM model for detecting fraud in credit card transactions. The effectiveness of our suggested hybrid CNN‐SVM model for detecting fraud in credit card transactions was tested using real‐world public credit card transaction data. The architecture of our hybrid CNN‐SVM model was developed by replacing the final output layer of the CNN model with an SVM classifier. The first classifier is a fully connected layer with softmax that is trained using an end‐to‐end approach, whereas the second classifier is a support vector machine that is piled on top by deleting the final fully connected and softmax layer. According to experimental results, our hybrid CNN‐SVM model produced classification performances with accuracy, precision, recall, F1‐score, and AUC of 91.08%, 90.50%, 90.34%, 90.41, and 91.05%, respectively.