Volume 14, Issue 4, October 2023
Prediction of Diabetes in Imbalanced Data Using Feature Computing and Hybrid Machine Learning Approach |
Pages: 1-10 (10) | [Full Text] PDF (643K) |
Amina Kalsoom, Amjad Hussain |
Department of Computer Science, University of Engineering and Technology, Pakistan
School of Systems and Technology, University of Management and Technology, Lahore-54700, Pakistan |
Abstract - Diabetes has emerged as a significant global health concern, affecting millions of individuals worldwide. Diabetes can have a significant impact on patient outcomes and healthcare costs, but early detection and prediction can considerably contribute to timely interventions, improving patient outcomes and saving healthcare costs. utilising the strength of machine learning algorithms, we offer a new method for diabetes prediction utilising the PIMA dataset. The PIMA dataset is a great tool for studying diabetes because of its comprehensive collection of clinical and demographic information. Our research aims to better understand the factors that put people at risk for getting diabetes and to create a prediction model that can accurately and quickly identify those people. We aim to extract interesting patterns and predictive indicators from the dataset by using a variety of machine learning techniques, such as classification algorithms and feature selection approaches. This paper aims to improve the model's performance, decrease its computational complexity, and increase its interpretability by carefully picking essential features. Our suggested method is shown to be effective in predicting diabetes from the PIMA dataset, as shown by experimental findings. In terms of early detection and risk assessment, our model shows promising predictive performance. Researchers, healthcare providers, and policymakers have a strong tool at their disposal with the combination of machine learning and the PIMA dataset, which will allow for the creation of more precise treatments, individualised treatment plans, and preventative measures. |
Index Terms - Diabetes, Healthcare, PIMA, ML, Detection, Prediction and Computational Complexity |
Simplistic Graphic User Interface Design For Rendering Blockchain Data (A Case Study of Educational Administrative Domain) |
Pages: 11-18 (8) | [Full Text] PDF (561K) |
Oluseyi Ayodeji Oyedeji, Ibiyinka Temilola Ayorinde |
Department of Computer Science, University of Ibadan, Nigeria |
Abstract - Blockchain data is characterized by encryption which is usually aimed at securing the data itself from getting into wrong hands. However, this limits the readability of the data when there is a need to. Clarity and understanding are essential indicators of a usable system as a complex but functioning system is of no value to the user if the data are not presented in a simple readable format. A Graphic User Interface (GUI) with a simplistic design is presented to solve this problem by rendering only necessary blockchain data while considering supplementary details that are usually hidden. 5 blockchain scenarios are created for this purpose. The activities and potential data involved are clearly stated. The raw blockchain data on a high level are recorded in form of JavaScript Object Notification (JSON). Attention is given to supplementary details from the raw blockchain data while translating them into graphical form. Results were presented in three different forms (raw blockchain data, complex GUI, and the simplistic GUI) to different categories of users and their comments were compared and analyzed. Since human computer interaction is an essential factor in user’s readability, the GUI rendered in this work is an enhancement to the blockchain technology as seen in the result which shows the simplistic GUI as the best understood interface even by a novice among the three. |
Index Terms - Blockchain, Human Computer Interaction, Interface and JavaScript Object Notification |
DDOS Attack Detection System for IoT Network Devices Using Machine Learning Techniques |
Pages: 19-28 (10) | [Full Text] PDF (498K) |
Khizra Mazhar, Junaid Arshad |
Computer Science Department, UET, Lahore |
Abstract - With the rapid growth of the Internet of Things (IoT) and the increasing prevalence of smart home devices, security concerns have become a critical issue. Among these concerns, distributed denial-of-service (DDoS) attacks pose a significant threat to the availability and functionality of smart home IoT devices. This paper focuses on the development of a machine learning-based approach for detecting DDoS attacks in smart home IoT devices. To achieve this goal, we employ a Logit Boosted algorithm that combines multiple weak classifiers to create a strong classifier for DDoS attack detection. Specifically, we utilize several variations of the Logit Boosted algorithm, including Logit-CBC, Logit-GBC, Logit-XGB, Logit-HGBC, Logit-LGBC, and Logit-ABC. Our experiments show promising results in terms of detection accuracy. The Logit-CBC algorithm achieves a detection rate of 98%, followed by Logit-GBC at 96%, and Logit-XGB at 91%. Although Logit-HGBC, Logit-LGBC, and Logit-ABC exhibit slightly lower detection rates at 78%, 81%, and 79% respectively, they still demonstrate the potential for effective DDoS attack detection. Furthermore, we evaluate the proposed approach using a real-world dataset of smart home IoT device logs. The results indicate the effectiveness of our Logit Boosted algorithms in accurately detecting DDoS attacks, thereby enhancing the security and resilience of smart home IoT networks. In conclusion, this paper presents a machine learning-based approach for DDoS attack detection in smart home IoT devices. The Logit Boosted algorithms, including Logit-CBC, Logit-GBC, Logit-XGB, Logit-HGBC, Logit-LGBC, and Logit-ABC, demonstrate varying levels of detection accuracy, highlighting the potential for effective defense mechanisms against DDoS attacks in smart home IoT environments. |
Index Terms - ML, DL, IoT Devices, DDoS, Security and Techniques |
Pages: - (7) | [Full Text] PDF (434K) |
Abstract - |
Index Terms - |
Pages: - (7) | [Full Text] PDF (655K) |
Abstract - |
Index Terms - |