Conference PPT on kl university for ug students

conference

1. INTERNATIONAL CONFERENCE FINTECH AND FINANCIAL INCLUSION “ EXPLORING AI AND MACHINE LEARNING SOLUTIONS IN REGULATORY TECHNOLOGY ” Mr. B V V S N MURTY 1 1 Department of Computer science and Engineering, Institute of Advanced Research, Gandhinagar, Gujarat Mr. P Srinivasa Praveen 2 2 Department of Statistics, A.B.N & P.R.R College of Science, Kovvur, Andhra Pradesh

2. Contents Abstract Introduction Literature Review Data Source Methodology Results Discussion Conclusion References

3. Abstract This paper examines the intersection of Artificial Intelligence (AI) and Machine Learning (ML) within the realm of Regulatory Technology ( RegTech ). RegTech , a burgeoning field aiming to streamline compliance processes and mitigate regulatory risks, has increasingly turned to AI and ML solutions to address complex regulatory challenges. Through a comprehensive review of literature and case studies, this paper explores various applications of AI and ML in RegTech , including risk assessment and monitoring, compliance automation, fraud detection and prevention, KYC/AML compliance, regulatory reporting, regulatory intelligence, and behavioral monitoring and surveillance. Additionally, the paper discusses the benefits, challenges, and future prospects of integrating AI and ML technologies into RegTech solutions, emphasizing the potential for enhanced efficiency, accuracy, and adaptability in regulatory compliance efforts. How effective are AI and ML algorithms in detecting and preventing fraudulent activities in financial transactions, and what factors influence their performance?

4. Introduction In recent years, the financial landscape has witnessed a surge in regulatory requirements and compliance obligations, driven by increasing scrutiny from regulatory bodies and a growing emphasis on transparency and accountability In response to these challenges, Regulatory Technology ( RegTech ) has emerged as a vital sector, leveraging technological innovations to facilitate compliance processes and manage regulatory risks effectively Artificial Intelligence (AI) and Machine Learning (ML) have emerged as game-changers in the RegTech domain, offering unparalleled capabilities in data analysis, pattern recognition, and automation This introduction sets the stage for exploring the convergence of AI and ML with RegTech , highlighting the transformative potential of these technologies in revolutionizing regulatory compliance We will delve into the various applications of AI and ML in RegTech , from risk assessment and compliance automation to fraud detection and regulatory reporting. Moreover, we will discuss the implications of integrating AI and ML solutions for enhancing efficiency, accuracy, and adaptability in navigating the complex regulatory landscape. By examining the benefits, challenges, and future prospects of AI and ML in RegTech , this exploration aims to shed light on the evolving paradigm of regulatory compliance in the digital age

5. Literature Survey " RegTech : Addressing the Challenges of Regulatory Compliance Using AI and Machine Learning" This review provides an overview of the challenges faced by organizations in meeting regulatory compliance requirements and explores how AI and machine learning technologies can be leveraged to address these challenges effectively. It discusses various applications of AI in RegTech , such as compliance monitoring, risk management, and regulatory reporting automation, and examines the benefits and limitations of implementing AI-driven solutions in regulatory environments. . “Machine Learning for Regulatory Compliance: Opportunities and Challenges" This paper reviews the existing literature on the use of machine learning techniques in regulatory compliance across different industries, including finance, healthcare, and telecommunications. "AI in RegTech : A Systematic Literature Review" This systematic literature review synthesizes findings from existing studies on the application of AI in regulatory technology. It categorizes the literature based on different aspects of RegTech , such as compliance monitoring, fraud detection, and regulatory reporting, and identifies common themes, trends, and research gaps in the field . "Regulatory Technology and Artificial Intelligence: A Review of Emerging Trends and Future Directions" This review examines emerging trends and future directions in the intersection of regulatory technology and artificial intelligence "The Role of AI and Machine Learning in Financial Regulatory Compliance: A Comprehensive Review" This comprehensive review provides an in-depth analysis of the role of AI and machine learning in financial regulatory compliance

6. Data Source The Bank data collected from the site” https:// www.kaggle.com/datasets/ jainilcoder / onlinepayment-fraud-detection?resource =download ” and Sample size(n) is 10000 Features Description Type.of.Transaction DEBIT, PAYMENT, TRANSFER, CASH-IN AND CASH-OUT Transaction.amount The transaction's amount expressed in local currency. oldbalanceOrg the client who initiated the transaction newbalanceOrig fresh balance following the transaction newbalanceDest Receiver of the new balance following the transaction. Please take note that there is no information available for clients who begin with Merchants. is.Fraud These are the transactions that the phone agents in the simulation made. The agents' fraudulent behavior in this particular dataset is focused on making money by seizing control of their customers' accounts, attempting to withdraw all of the money by moving it to another account, and then cashing out of the system.

7. Methodology The methodology for evaluating the effectiveness of AI and ML algorithms in detecting and preventing fraudulent activities in financial transactions involves several key steps. First, a diverse dataset of financial transactions is collected, encompassing both legitimate and fraudulent instances. This dataset is then pre-processed to handle missing values, outliers, and inconsistencies before being split into training, validation, and test sets. Feature engineering is performed to extract relevant features and create additional ones that may enhance fraud detection performance. Next, suitable AI and ML algorithms, such as logistic regression is selected and trained using the training dataset. In this model has predicted variable is fraudulent activates and exploratory variables are transaction time, Type of Transaction, Transaction amount , Customer starting the transaction, old balance Origen, new balance Origen, Transaction.ID, old balance Destination, new balance Destination. The logistic Regression equation is

8. Results TRANSACTION TYPE FREQUENCY PERCENTAGE CASH_IN 2179 21.79% CASH_OUT 3525 35.25% DEBIT 77 0.77% PAYMENT 3402 34.02% TRANSFER 817 8.17% TOTAL 10000 100%     Table 1: Transaction type Fig 1: Pie Diagram for the Type of Transaction

9. Results Table 2: Transaction Time : TYPE OF TRANSACTION MEAN TRANSACTION TIME CASH_IN 26.80644125 CASH_OUT 26.67753539 DEBIT 25.83853441 PAYMENT 27.3405346 TRANSFER 27.19378004 Table 3: Transaction amount TYPE OF TRANSACTION FREQUENCY CASH_IN 1399284 CASH_OUT 2237500 DEBIT 41432 PAYMENT 215195 TRANSFER 532909

10. Results Fig 2: Bar Chart for Transaction mount for Each type of Transaction 2500000 2237500 2000000 1399284 1500000 1000000 532909 500000 215195 41432 CASH_IN CASH_OUT DEBIT PAYMENT TRANSFER

11. Results Table 4: Frequency table for Is Fraud: CATEGORY FREQUENCY PERCENTAGE NOT FRAUD 1047433 99.89% FRAUD 1142 0.11% TOTAL 1048575 100% Fig 3: Bar chart for Fraud of transactions

12. Results Table 5: Frequency table for Type of Transaction and Number of Fraud transactions TRANSACTION TYPE NUMBER OF FRAUD TRANSACTIONS CASH_IN CASH_OUT 4107 DEBIT PAYMENT TRANSFER 4107 TOTAL 8214    

13. Results Fig 4: Bar chart for Number of Fraud transaction in each type of transaction Fig 5 : Amount per transaction by Fraud:

14. Results Fig 5 : Amount per transaction by No Fraud:

15. Results Table 7 : Correlation Matrix   oldbalanceOrg newbalanceOrig amount isFraud oldbalanceOrg 1.000 0.997 0.125 -0.013 newbalanceOrig 0.997 1.000 0.091 -0.034 amount 0.125 0.091 1.000 0.137 isFraud -0.013 -0.034 0.137 1.000

16. Results Table 8: Logistic Regression model output   COEFFICIENT S. ESTIMATE COEFFICIENT S. STD..ERROR Z- VALUE P-VALUE (INTERCEPT) -3.916 0.305 -12.82 0.000 TYPE 0.761 0.117 6.49 0.000 AMOUNT -0.0002 0.000 -5.25 0.000 OLDBALANCEORG 0.0002 0.000 5.73 0.000 NEWBALANCEORIG -0.029 0.026 -1.10 0.271 From output Logistic regression model for binary variable Is Fraud

17. Results Table 8: Logistic Regression model output   COEFFICIENT S. ESTIMATE COEFFICIENT S. STD..ERROR Z- VALUE P-VALUE (INTERCEPT) -3.916 0.305 -12.82 0.000 TYPE 0.761 0.117 6.49 0.000 AMOUNT -0.0002 0.000 -5.25 0.000 OLDBALANCEORG 0.0002 0.000 5.73 0.000 NEWBALANCEORIG -0.029 0.026 -1.10 0.271 From output Logistic regression model for binary variable Is Fraud

18. Results Table 9: Confusion Matrix Actual value Predicted value Fraud Not Fraud Fraud 18 31 Not Fraud 2 6297 Accuracy Score = 99.48% Interpretation of Accuracy Score: It m eans that approximately 99.48% of the instance in this dataset were correctly classified by the model. A high accuracy score suggests that this model is making accurate predictions for the majority of the instances.

19. Discussion Intercept (𝛽0 = −3.916): In the above model, when all other predictors are zero, the expected value of the dependent variable is approximately - 3.916 Coefficient of Transaction type(𝛽1 = 0.761): For a one-unit increase in the "Type" variable, the expected change in the dependent variable may Fraud because it is an increase of approximately 0.761. So it very near to 1 Coefficient of Amount (𝛽2 = −0.0002): one-unit increase in the "Amount" variable, the expected change in the dependent variable is a decrease of approximately 0.0002 . Coefficient of Old Balance (𝛽3 = 0.0002): For a one-unit increase in the " OldBalanceOrg " variable, the expected change in the dependent variable (is Fraud) is an increase of approximately 0.0002 Coefficient of NewBalanceOrig (-0.029): For a one-unit increase in the " NewBalanceOrig " variable, the expected change in the dependent variable (Is Fraud is a decrease of approximately 0.029 units, holding all other variables constant . The coefficients provide insights into the impact of each predictor variable on the likelihood of fraud.

20. Conclusion The explanatory variables such as transaction type (Type), transaction amount(amount), and client initiating the transaction ( OldbalanceOrig )are identified as significant predictors for binary variable indicating fraud, as their p-values are below the significance threshold of 0.05. However the variable representing the fresh balance following the transaction ( NewbalanceOrig ) is found not to be a significant predictor for fraud, with p-value of 0.271exceeding the significance level Overall this logistic regression model offers valuable insights to the determinants of fraudulent activity in financial transactions. By examining factors like transaction type, amount and balance the model demonstrates the ability to accurately predict the likelihood of fraud. Nevertheless, it’s important to recognize the model limitations. Including its reliance of relationship and the potential existence of unobserved variables

21. References Bhattacharyya, S., & Jha , D. (2019). "An Ensemble Model for Fraud Detection in Financial Transactions Using Machine Learning Techniques." International Journal of Computational Intelligence Systems , 12(1), 794-807 . Phua , C., Lee, V., Smith, K., & Gayler , R. (2016). "A Comprehensive Survey of Data Mining-Based Fraud Detection Research." ArXiv Preprint arXiv:1609.06686 . Dal Pozzolo , A., Boracchi , G., Caelen , O., & Alippi , C. (2015). "Learned Lessons in Credit Card Fraud Detection from a Practitioner Perspective." Expert Systems with Applications , 41(10), 4915-4928 . Cortes, C., & Vapnik , V. (1995). "Support-Vector Networks." Machine Learning , 20(3), 273-297 . Ribeiro , M. T., Singh, S., & Guestrin , C. (2016). "Why Should I Trust You?" Explaining the Predictions of Any Classifier. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , 1135-1144. Hossain , M. S., Muhammad, G., & Alhossaini , S. M. (2017). "A Comparative Study of Credit Card Fraud Detection Techniques: Data Mining Approach." International Journal of Data Mining & Knowledge Management Process, 7(2), 51-64.

22. T h a n k y o u