With the new trend of Online Shopping and Online Platforms for transactions, the number of Credit Card based transactions increased tremendously. However, there have been a lot of cases where illegal use of Debit/Credit Cards for making Fraudulent Transactions. Credit card companies have been paying a lot of attention to providing the best service for their customers by having process enhancements and pro-actively looking into transactions before making them through. Global financial losses related to payment cards are estimated to reach $34.66 billion in 2022, according to The Nilson Report, a newsletter that tracks the payment industry. Related to the negative impacts of credit card fraud activities, and financial and product losses, it’s easy for merchants and users to feel victimized and helpless. Machine Learning Models can work well in detecting such Fraudulent actions when they are trained on a large quantity of historical data and then fine-tuned depending on validation and evaluation metrics. Speaker(s): Priyanka Singh, Devy Ratnasari, Gopika Shaji, Oluwole Ayodele, Saurav Bisht, Toronto, Ontario, Canada, Virtual: https://events.vtools.ieee.org/m/312336

Text Summarization is a technique for generating a concise and precise summary of voluminous texts while focusing on the sections that convey useful information without losing the overall meaning. It aims to transform lengthy documents into shortened versions, which could be difficult and costly to undertake if done manually. With the current explosion of data circulating in digital space, primarily unstructured textual data, there is a need to develop tools that allow people to get insights from them quickly. In situations where it is essential to keep track of what is being spoken, such as during an online lecture, taking notes is a popular activity used by many. The art of notetaking does not involve making notes of every single word that is spoken but comprehensive outlines of what is discussed. The key to good notetaking lies in making concise yet informative summaries. In this seminar, we will be discussing how we have tried to address the difficulties of notetaking by building an application that produces notes based on transcripts generated by the Automatic Speech Recognition (ASR) technology of the meeting platforms. We experimented with six summarization models for this application, including transformer-based models pre-trained on large corpora. The datasets used for this application are the transcripts dataset acquired from online meeting platforms and the Extreme Summarization (XSum) dataset. We evaluated the models using Rouge metrics (Rouge-1, Rouge-2, and Rouge-L) and selected the best-performing model as the final model. We have built a bot that utilizes Telegram’s API and shares the generated summaries via group chat with the users. Speaker(s): Manoj Varma Alluri, Navaneeth Jawahar, Sharath Kumar Prabhu, Jeel Jani, Shravya Sandupata Toronto, Ontario, Canada, Virtual: https://events.vtools.ieee.org/m/312337

The rise of digitalization and the advent of social media and e-commerce have generated an abundance of data than before. Natural Language Processing (NLP) is a significant branch of artificial intelligence that helps the machine interpret human languages and perform the desired task by analyzing the semantics, content, and pattern. Sentiment analysis is the most common technique in Natural Language Processing used to determine the underlying sentiments of a text. This technique is currently in place for different Business Organizations to analyze their brand’s market value, brand reputation, and customer perception of new brand/new change. Businesses use social media channels to cater to their customer service, and people use social media to express/share their wide range of opinions or experiences about a product/brand. These opinions and experiences reflect the real-time sentiments of a customer. Sentiment analysis will help businesses designing an effective marketing campaign, better customer satisfaction, boost sales, help improve customer experience, understand customer perception to change and the brand’s market reputation. The customer views expressed on Twitter, Facebook, and other online forums are forming the base of customer strategy for brands worldwide. Businesses are opting to shift their traditional customer feedback analysis method to text classification since people prefer to post the genuine reviews on the internet. Analyzing the underlying sentiments in the text will help the business to understand their customers' voices and their brand reputation in the market in real-time. Sentiment analysis will help the businesses designing an effective marketing campaign, better customer satisfaction, boost sales, help improve customer experience, understand customer perception to change and the brand’s market reputation. Twitter sentiment analysis aims to classify text into positive/negative based on its underlying semantics. Speaker(s): Akhil Mathew, Anmol Wadera, Deepan Ellenti Padmanabhan, Saketh Vemula, Sivaramakrishna Malakalapalli Register: https://events.vtools.ieee.org/m/312338

The research goal is to implement different machine learning algorithms to detect any DDoS (Distributed Denial of Service) attacks using the UNSW-NB15 dataset. We started by going through the data description and finding null values in our features. After that we dropped the ‘id’ column. We have used the UNSW-15 dataset for AI-based DDOS detection systems. The UNSW-15 dataset has a hybrid of the real modern normal and the contemporary synthesized attack activities of the network traffic. It contains different attacks, including DoS, worms, Backdoors etc. The raw network packets of the UNSW-NB 15 datasets are created by the IXIA Perfect Storm tool in the Cyber Range Lab of the Australian Centre for Cyber Security (ACCS) for generating a hybrid of real modern normal activities and synthetic contemporary attack behaviours. We incorporated different feature selection methods for dropping insignificant features followed by the implementation of 6 classification algorithms, namely Naive Bayes, Random Forest, Decision Tree, KNN, Logistic Regression and SVM. Speaker(s): Minu Ahlawat, Dwij Dua, Megha Garg, Taxil Savani Register: https://events.vtools.ieee.org/m/312339

Cancer ranks as a leading cause of death and an important barrier to increasing life expectancy everywhere. According to available data, lung cancer contributes the most to cancer deaths. Also, according to available data, those diagnosed early have a 50 percent chance of survival over those diagnosed with late-stage cancer. It means that early detection is paramount to the survival of a lung cancer patient, leading to a reduction in the number of cancer deaths. We, therefore, evaluated six different machine learning algorithms to see which one performed optimally in accurately predicting the level of lung cancer development in a patient. We considered various parameters when choosing the dataset for this evaluation as the pathogenesis of lung cancer involves a combination of intrinsic factors and exposure to environmental carcinogens. We also considered varying the features in our data, categorizing them under diagnostic risk factors (age, gender, alcohol use, air pollution, balanced diet, obesity, smoking, passive smoker) and symptoms (fatigue, weight loss, shortness of breath, swallowing difficulty, frequent cold, dry cough) and the inferences we drew from this indicated that those that have the symptom features prior to diagnosis had the highest chance of being diagnosed with a high level of cancer. The final results of our evaluation showed that the best levels of predictions on new data were achieved by optimized Random Forest, KNN, and SVM models. Speakers: Rakesh Pattanayak, Chisom Nnabuisi, Dhruv Mistry, Kar Chun Kan, Shanuka Rathnayake Register: https://events.vtools.ieee.org/m/312340