deep learning and knowledge graph for financial loss framework

knowledge graph for financial loss framework

1. A deep Learning and knowledge graph based framework for detection of financial loss cases in online employment scams End-Sem Presentation Presented By: Shikhar Mishra En. ID: SAU/CS(M)/2022/02 Supervisor: Prof. Muhammad Abulaish Laboratory for Data Science and Analytics (LDSA), Department of Computer Science, South Asian University

2. Introduction and recap Data, information and knowledge graph. Knowledge Graph: Nodes (entities), edges (relationship) and triplets. Some well known examples: DBpedia , Freebase, etc. Applications in AI (Artificial Intelligence) Systems: Recommender System, Question Answering System, etc. Application fields: Education, Scientific Research, Social Networks, etc. Knowledge Graph Technologies: Knowledge Acquisition, Knowledge Graph Completion, Knowledge Fusion, etc. Problems related to knowledge graph technologies. 1

3. Introduction and recap (cont.) Implementation on Dbpedia endpoint and Jupyter Notebook. INPUT - select * where {?athlete rdfs:label “Cristiano Ronaldo”@ en ; dbo:birthPlace ?Birth. ?Birth a dbo:City ; rdfs:label ? CityName ; dbp:region ?Region. filter(lang(? CityName )in(‘ en ’, ‘ ja ’, ‘ fr ’))} OUTPUT - 2

4. INPUT OUTPUT 3 Introduction and recap (cont.)

5. Machine learning Machine learning (ML) is the study of computer algorithms that can improve automatically through experience and by the use of data. It is seen as a part of artificial intelligence or branch of artificial intelligence. Machine learning algorithms build a model based on sample data. Such models are build to make predictions or decisions without being explicitly programmed to do so. Here, model is trained on training dataset and tested on test dataset. Test dataset evaluates the performance of the model and ensures that the model can generalize well with the new or unseen dataset 4

6. Categories of ml Supervised machine learning Unsupervised machine learning Semi-supervised learning Reinforcement Learning 5

7. Neural networks A neural network is a network of artificial neurons or nodes. Neural networks  are artificial systems that were inspired by biological neural networks. It has a huge number of linked processing nodes. Good at recognizing patterns. Artificial neural networks are used for solving artificial intelligence (AI) problems. These systems learn to perform tasks by being exposed to various datasets and examples without any task-specific rules. Important role in applications including natural language translation, image recognition, etc. 6

8. Types of neural network Multilayer Perceptron (MLP): A type of feedforward neural network with three or more layers, including an input layer, one or more hidden layers, and an output layer. Convolutional Neural Network (CNN): A neural network that is designed to process input data that has a grid-like structure, such as an image. It uses convolutional layers and pooling layers to extract features from the input data. Recursive Neural Network (RNN): A neural network that can operate on input sequences of variable length, such as text. Recurrent Neural Network (RNN): A type of neural network that makes connections between the neurons in a directed cycle, allowing it to process sequential data. Sequence-to-Sequence (Seq2Seq): A type of neural network that uses two RNNs to map input sequences to output sequences, such as translating one language to another. 7

9. Deep learning Deep learning (DL) is a branch of machine learning that is made up of a neural network with three or more hidden layers: Input layer:  Data enters through the input layer. Hidden layers:  Hidden layers process and transport data to other layers. Output layer:  The final result or prediction is made in the output layer. Neural networks inspired by human learning try to model by digesting and analyzing massive amounts of information, also known as training data. They perform a given task with that data repeatedly, targeting to improve accuracy each time. 8

10. Some Widely used dl models in NLP LSTM (Long Short Term Memory) LSTMs are an extension of recurrent neural networks (RNNs). It excels at storing long-term dependencies . It removes vanishing gradient problem which is during the training of deep neural networks the gradients that are used to update the network become extremely small or "vanish" as they are backpropogated from the output layers to the earlier layers. GRU (Gated Recurrent Unit) It is a type of recurrent neural network (RNN) that can process sequential data such as text, speech, and time-series data. Basic idea is to use gating mechanisms to selectively update the hidden state of the network at each time step. Bi-LSTM In this, input flows in both directions, and it’s capable of utilizing information from both sides. It adds one more LSTM layer, which reverses the direction of information flow. 9

11. NLP and some common techniques Natural Language Processing (NLP) is a subfield of artificial intelligence that deals with the interaction between computers and humans in natural language. Some common techniques used in NLP include: Tokenization: the process of breaking text into individual words or phrases. Named entity recognition: the process of identifying and categorizing named entities, such as people, places, and organizations, in text. Sentiment analysis: the process of determining the sentiment of a piece of text, such as whether it is positive, negative, or neutral. 10

12. Dataset (Employment Scam) Source of the dataset is BBB Scam Tracker¹, an online platform where consumers and businesses report scams. Extracted around 10k employment scam complaints of USA from 18 th Nov, 2022 to 24 th Nov, 2023. Dataset consists of scam id, textual description, dollar value of any loss, scam location, scam’s reported date, scammer’s location, mail id of scammer, contact number and website url . 11 ¹https://www.bbbmarketplacetrust.org/

13. 12

14. 13

15. Experimentation We performed experiment on two neural network models named multilayer perceptron (dense classifier) and LSTM 15

16. Pseudocode 14

17. 15

18. Output (Result of both models on test data) 16 Precision Recall F1-Score Precision Recall F1-Score Dense Classifier LSTM

19. Output (Result of both models on test data) 17 Accuracy

20. Future works This is a case of class imbalance where the case of financial losses are in minority. We will utilize the potential of knowledge graph and deep learning techniques to improve the performance of the classifier. 18

21. References 19 Peng, C., Xia, F., Naseriparsa , M., & Osborne, F. (2023). Knowledge graphs: Opportunities and challenges.  Artificial Intelligence Review , 1-32. Auer, S., Bizer , C., Kobilarov , G., Lehmann, J., Cyganiak , R., & Ives, Z. (2007, November). Dbpedia : A nucleus for a web of open data. In  international semantic web conference  (pp. 722-735). Berlin, Heidelberg: Springer Berlin Heidelberg. Goyal, N., Mamidi , R., Sachdeva, N., & Kumaraguru , P. (2023, January). Warning: It’sa scam!! Towards understanding the Employment Scams using Knowledge Graphs. In  Proceedings of the 6th Joint International Conference on Data Science & Management of Data (10th ACM IKDD CODS and 28th COMAD)  (pp. 303-304). Grant-Smith, D., Feldman, A., & Cross, C. (2022). Key trends in employment scams in Australia: What are the gaps in knowledge about recruitment fraud?.  QUT Centre for Justice Briefing Papers ,  21 . Goyal, N., Mamidi , R., Sachdeva, N., & Kumaraguru , P. (2023, January). Warning: It’sa scam!! Towards understanding the Employment Scams using Knowledge Graphs. In  Proceedings of the 6th Joint International Conference on Data Science & Management of Data (10th ACM IKDD CODS and 28th COMAD)  (pp. 303-304). Grant-Smith, D., Feldman, A., & Cross, C. (2022). Key trends in employment scams in Australia: What are the gaps in knowledge about recruitment fraud?.  QUT Centre for Justice Briefing Papers ,  21 . www.wikipedia.org https://medium.com/virtuoso-blog/dbpedia-basic-queries-bc1ac172cc09 www.geeksforgeeks.org

22. THANK YOU