Want to make interactive content? It’s easy in Genially!

Over 30 million people build interactive content in Genially.

Check out what others have designed:

Transcript

Using SVM algorithm in order to enhance the Financial Alert Management to reduce false positive and increase true positive transactions

Department of Information Engineering, Computer Science and Statistics (DIAG).

Zahra Rezaei Estakhroueiyeh 1959632

Superviser: Professor Luca Iocchi

March 2024

Master of Artificial Intelligence and Robotics

Index

1.

Summary

2.

Introduction

3.

Quality check on Alert Evaluation

4.

Support Vector Machine (SVM)

5.

6.

7.

Libraries

Conclusions

Bibliography

1. Summary

Summary:

  • The role of artificial intelligence (AI) in enhancing business efficiency and risk management
  • How Machine Learning enables:
smarter decision-making, faster processes, cost reduction, improved accuracy, enhanced customer experience
  • Basic goals about Quality check on Alert Evaluation
  • Introducing Support Vector Machine
  • AI-powered solutions are enhancing efficiency, accuracy, and innovation.
  • Effective alert management is paramount.
  • harnessed SVM within the Palantir platform
  • Area Under Curve (AUC) value of 98.2 percent

2. Introduction

Introduction:

3. Quality check on Alert Evaluation

Quality Check on Alert Evaluation

  1. Evaluating Alert Quality
-Incoherence Verify -Commentary Length Verify -Misalignment with GOR -Sigra Cubana Case Verify -Same Match Closed Differently -FP Cases with >=2 Names Matching

3. Quality check on Alert Evaluation

Quality Check on Alert Evaluation

2. Enhancing Risk Analysis and Reporting -Definition and Implementation of Enhancement Logic - Creation of Code Repository - Creation of User-Friendly Report for Risk Analysis: -Testing and Delivery of Report:

3. Quality check on Alert Evaluation

Quality Check on Alert Evaluation

3. The steps involved the FS Risk Analysis and Quality Check process -FS Risk Analysis -FS Training: -FS Continuous Monitoring (TBE - To Be Evaluated)

4.Support Vector Machine

SVM propertoes:

  • Belongs to the class of binary classifiers
  • Well-suited for solving classification problems in high-dimensional spaces
  • Handling complex decision boundaries
  • The model can analyze various features and patterns associated with financial transactions
  • Detect and classify patterns

The specific implementation within the Palantir platform :

  • Dataset Selection
  • Label Conversion
  • Feature Selection
  • Data Splitting
  • Model Training
  • Performance Evaluation
  • Visualization
  • SVM Model Training
  • Pattern Learning
  • Prediction Application
  • Comparison with Actual Labels
  • Decision Boundary

5. Librariesn

Libraries:

  • from pyspark.ml.feature import StringIndexer, VectorAssembler
  • from pyspark.ml.classification import LinearSVC
  • from pyspark.ml import Pipeline
  • from pyspark.ml.linalg import DenseVector
  • from pyspark.sql import functions as F
  • from pyspark.ml.evaluation import BinaryClassificationEvaluator
  • from sklearn.calibration import Calibrated-ClassifierCV
  • import numpy as np

8. Development

Results:

98.2%

AreaUnder Curve
  • The results were significant, with the SVM model achieving an impressive AreaUnderCurve (AUC) value of 98.2 percent, during testing.
  • Introducing SVM into financial alert management, marks a
- significant advancement, against financial fraud - increasing more satisfaction to customers

6..Conclusions

Conclusions

  • Internship Insights: educe false positives and increase true positive detections—critical for preventing fraud and satisfying customers
  • Impressive Results: Area Under the Curve (AUC) value of 98.2 percent, showcasing its efficacy in alert management.
  • Operational Transformation: transform financial institution operations, leading to improved security, operational efficiency, and customer trust

LMoraes, Daniel and Wainer, Jacques and Rocha, Anderson, 2016, Low False Positive Learning with Support Vector Machines, https: //www.researchgate.net/publication/298427125_Low_False_Positive_ Learning_with_Support_Vector_Machines/citation/download

Daniel Moraes, Jacques Wainer, Anderson Rocha, July 2016, Low false positive learning with support vector machines https://www.sciencedirect.com/ science/article/abs/pii/S1047320316300116

LGiancarlo Cesi (Head Of Global Regulatory Track), Email Address: Giancarlo.Cesi@unicredit.eu

  • My supervisor/boss at UniCredit Bank:
  • Paper

10. Bibliography

Bibliographic references

  • Paper

LJanuary to July (2023). Internal Policies and Procedures. Milan, Italy: UniCredit Bank.

  • UniCredit Bank (Milan)

Thank you!

Any questions?