Prabakaran R

Bio: Prabakaran R is an academic researcher. The author has contributed to research in topics: Credit score & Conversation. The author has an hindex of 1, co-authored 3 publications receiving 3 citations.

SecrecyProtector: A Novel Data Analytics based credit score management system

Abstract: This work gives an account of the Credit Score web service application and the primary purpose of a credit score is to help lenders assess individuals' risk of not repaying a loan. Credit scoring assessment, despite the fact that a moderately new idea in the Indian money related business sector, have increased wide acknowledgment among financial specialists. In the meantime, easy-going and narrative confirmation recommends that there are worries among speculators and controllers about the execution of rating offices in India. This paper looks at financial specialists' mindfulness, discernment, understanding level and use of Credit scoring assessment through a poll-based example overview covering individual and additionally institutional speculators. We find high dissemination of rating use among all class of financial specialists, however, there is a recognizable upsetting with the dependability of appraisals, inclination of ensuing minimizing and opportuneness of rating reconnaissance. The review additionally uncovers that the institutional financial specialists have predominant information and comprehension about evaluations than individual speculators. In this way, the review underlines the requirement for rating offices to take a shot at instructing the basic speculators to engender appropriate comprehension and use of Credit

Towards Building A Neural Conversation Chatbot Through Seq2Seq Model

Abstract: Improvements in computation and processing power paved a way for Machine learning to be applied more efficiently in real-time and in a lot of applications. In which most prominent area is Natural Language Processing and Natural Language Understanding, which helps the computer to process and understands the natural language used by people. Thanks to deep learning models and architectures which made this process of making the system process and understand natural language, which makes the system more intelligent. Chatting agent’s AKA-Chatbot is one of the major use cases of Natural Language Processing and Natural Language Understanding, which can be used in different domains to engage customers and provide a response to customer’s queries. Though many chatbots use a retrieval-based model with the recent advancement of Deep Learning, we in this work use Neural Networks to train a chat model with a question and answer datasets that make models understand the patterns in it and behave intelligently. Here we build a domain-specific generative chatbot using Neural Networks to train a conversational Model which reads the pattern of data and reply answer when a new question is asked. Finally, we conclude by validating how relevant the response generated by the model to test data or test question and provide a further area of improvements to make the system more efficient and intelligent.

Skin Cancer Detection using Deep Learning

Abstract: Identifying melanoma at the early stages of diagnosis is imperative as early detection can exponentially increase one’s chances of cure. The paper first proposes a literature survey of multiple methods used for performing skin cancer classification. Our methodology consists of using Convolutional Neural Network (CNN) to identify and diagnose the skin cancer using the IS IC dataset containing 2637 images. The proposed model gives an accuracy of 88% for classifying the training dataset as either benign or malignant.

Intruder Detection System using IoT with Adaptive Face Monitoring and Motion Sensing Algorithm

Abstract: Internet of Things (IoT) conceptualizes the possibility of distantly interfacing and checking things through the web. At the point when it includes our home, this thought is frequently suitably fused to shape it more brilliant, more secure and programmed. In this paper, a novel security approach algorithm for detection and recognition for home automation has been proposed with the assistance of IoT. The results show that by combining the CCTV and Motion sensor, the detection of intrusion is more efficient. This combination helps to reduce the blind spot. And a face recognition system to differentiate between the homeowner and others. Then alerting the homeowner about the intrusion.

A Robust Pipeline Approach for DDoS Classification using Machine Learning

Abstract: Remote and edge devices have less security features that are easily exploited by hackers. The security of businesses in major domains depends on the security features the infrastructure has to offer. Major breaches have been reported over the past years which have led to compromise of hidden data. DDoS attacks have been a major trend which has brought down many devices using similar techniques. Major vulnerabilities have been found in IoT systems which presents an open door for hackers. To address the upcoming trends in early vulnerabilities detection, a standard predictive model of DDoS attacks needs to be implemented. In this paper we propose a robust pipeline for DDoS classification and the performance of the models are calculated against the metrics such as precision, recall and f1-scores. After evaluating various machine learning models, the XGboost algorithm works well on our data set with an accuracy score of 99% outperforming other models.