Asian Journal of Convergence in Technology
Not a member yet
868 research outputs found
Sort by
Comparative Evaluation of Predictive Models on Kidney, Lung Cancer and Heart Disease
Get PDFThis study supports advances in machine learning to improve early detection and treatment planning for lung cancer, cardiovascular disease, and kidney disease. We compare traditional models such as decision trees and logistic regression with complex techniques such as support vector machines, random forests, and KNN and evaluate them on publicly available data. This hybrid approach uses random forest and decision tree classifiers, leveraging adaptive learning to improve model accuracy. Results showed high prediction accuracy for kidney disease and lung cancer , while prediction accuracy for heart disease was average . This difference indicates the need for better work and more information. Future studies will focus on improving cardiovascular models, addressing data uncertainty, and integrating predictive models into clinical practice to support early diagnosis and personalized treatment to improve patient outcomes. This study demonstrates the potential for machine learning to have a major impact on diagnosis and patient management
A Comprehensive study on Satellite Image Super-resolution using Diffusion and GAN based model
Get PDFObject detection and feature extraction from satellite images is a crucial step while using satellite images for purposes like navigation, urban planning, weather monitoring, etc. While deep learning approaches are too common for object detection task, but when the satellite images are of low quality, the small objects are missed by detection model due to their size and visibility issue. In this paper we propose a study of two broad areas of Generative AI models namely GANs and Diffusion model and their ability to handle the low-resolution images to improve overall detection problem. We train SRGAN and Diffusion based super-resolution model on custom real-time datasets and present a comprehensive performance evaluation and analysis. We found that Diffusion model increased the object detection rate by almost 130% when compared with Raw image object detection
Automated Brain Tumor Detection Using Deep Learning and Flask Web Application
Get PDFBrain tumors are an important health issue worldwide, and timely diagnosis is often needed for effective treatment strategies. With advances in deep learning techniques, automated tumor detection systems have emerged as promising tools to help radiologists perform more accurate diagnoses In this paper we present a comprehensive analysis of the brain presenting a tumor detection system developed using deep learning models integrated with the Flask web application. We discuss the program design, implementation, and performance evaluation, and highlight potential impacts on health care delivery
Use of Machine Learning Application for Business Perspective
Get PDFCustomer segmentation plays a crucial role in understanding customer behaviour and tailoring marketing strategies. This project focuses on using K-Means clustering, a popular unsupervised machine learning algorithm, for customer classification based on their purchasing behaviour. The objective is to develop a customer segmentation model that can effectively group customers into distinct clusters to facilitate targeted marketing efforts.
The project begins with the collection of a fictitious e-commerce dataset consisting of 5000 customers with their purchase history. The dataset includes features such as customer ID, age, gender, annual income, and spending score. Data preprocessing techniques are applied to handle missing values and standardize the data, ensuring accurate and meaningful analysis.
Feature extraction involves selecting relevant features from the dataset, including age, gender, annual income, and spending score. These features provide valuable insights into customer behaviour and serve as the basis for customer segmentation.
The K-Means clustering algorithm is employed to classify customers into distinct clusters based on their purchasing behavior. The algorithm partitions the customers into K clusters by minimizing the sum of squared distances between the customers and their respective cluster centers. The optimal value of K is determined using the elbow method, a visual technique that identifies the point of maximum curvature in the sum of squared distances plot.
The effectiveness of the K-Means clustering model is evaluated using the Silhouette score. This score measures how well each customer fits into its assigned cluster, with values ranging from -1 to 1. A higher Silhouette score indicates better cluster cohesion and separatio
Quick Sort Optimized for Non-decreasing Data set
Get PDFThe Sorting Algorithm proposed by C.A.R Hoare in 1961 by name Quick sort, which is popularly known for being the fastest sorting algorithm. Quick sort is still being practiced in the field of computers systems and its applications. The Algorithm whose efficiency to sort random data set is represented in asymptotic notation as O(n log2 n) and when quick sort algorithm has a input data set which is already Ordered then it takes a quadratic execution time which is considered as a worst case performance and this behavior is represented in asymptotic notation as O(n2). The worst case performance is due the scan over heads which occur over the pre-sorted data set, in other words the partitioning gets skewed due to recursive calls and hence results in a quadratic complexity. This research paper presents an algorithm which minimizes a worst case execution time making it linear when the input list is in non-decreasing order. The paper describes how the improvements are accommodated in the existing quick sort. A priori analysis of proposed algorithm for different cases is made along with a proof of correctness. Later the algorithm is verified for its correctness and asymptotic performance. The algorithm is implemented using C++ and also we have compared with other popular quick-sort version
ECHO: Empowering Children’s Healthcare with Humanoid Empathy
Get PDFIndividuals with Autism Spectrum Disorder (ASD) face multifaceted challenges in social interaction and communication, necessitating innovative approaches to support their unique needs. Current statistics highlight the pressing need for effective interventions, with ASD prevalence rates continuing to rise globally. However, existing solutions often encounter limitations, such as subjective assessments and lack of personalized approaches. To address these challenges, this paper presents a comprehensive review of studies to assess and support individuals with ASD. By synthesizing findings from diverse studies utilizing various robotic platforms, including Robotis Mini, Romo, CommU, RobotParrot, Zeno, and ONO robots, this review elucidates the potential of robotics in facilitating accurate assessment, personalized intervention, and enhanced engagement for individuals with ASD [1]. Furthermore, the incorporation of advanced technologies, including multimodal data analysis and real-time gesture recognition algorithms, underscores the interdisciplinary nature of this research domain. While promising, the implementation of robotics in ASD intervention is not without drawbacks, including technical limitations and ethical considerations. Through ongoing exploration and innovation, robotics holds the potential to revolutionize the landscape of ASD support, fostering greater inclusivity, empowerment, and quality of life for individuals across the autism spectrum [2]. We have been designed a robot as a diagnostic toy for mentally challenged children, integrating both hardware and software components to facilitate interactive and engaging experiences aimed at assessing behavioral, and social-emotional skills while offering support and companionship in therapeutic settings. Advanced features such as face recognition, emotion detection, object and colour recognition, teaching modules, and comprehensive reporting systems drive foster learning and development, and empower children to reach their full potential in a supportive and engaging environment. This innovative approach aims to enrich the lives of mentally challenged children, fostering positive outcomes through tailored interventions and interactive experiences
Image Pre-Processing Detection Using Deep Reinforcement Learning
Get PDFReinforcement learning is a popular domain of Machine learning being actively used in many industries. With proper reward function and problem formulation, a task can easily be solved using Reinforcement learning.
Computer vision domain often face the problem of noisy data. Even when the model is trained on good quality data, during inference, the noisy data can create problem and thus causing model to fail. An essential step to overcome the problem of bad data is pre-processing of data, but pre-processing of image data is itself a complex problem and requires subject matter experts to decide which kind of preprocessing should be applied on a given image for a particular compute vision task.
To solve this problem of choosing correct pre-processing for a given images, we have proposed a novel approach of automation of image pre-processing using deep reinforcement learning. This approach is developed and tested for one of the most popular problems in image data, which is noise in images, while it has also shown potential how it can change the whole scenario of machine learning models being applied in the field of computer vision
Exploring the Potential of Q-Learning Offers a Promising Pathway towards Achieving Artificially Intelligent Driving Capabilities
Get PDFThis research focuses on Artificially Intelligent driving techniques that are being used to train several machinelearning models to achieve complete human-like driving skills. Artificially Intelligent driving consists of training a machine to drive on a provided path to any vehicle (a car in this case) while simultaneously following all the traffic routes, providing passenger comfort and vehicle and passenger safety. In this research, since most of the available artificially intelligent driving models are set to work upon a predetermined path provided by the user and can only follow that path, I intend to further this model by providing a completely random path to the model and then evaluate its efficiency, the resources it requires to complete its whole path to training the model so that it can adapt to the randomly provided path with much faster speed and more accuracy as compared to traditional Artificially Intelligent vehicle driving models
CareerCraftML: Smarter Studying with AI-Powered Tools
Get PDFIn today's competitive job market, leveraging advanced technology to enhance job application processes and interview preparedness is crucial. This research paper presents a comprehensive system integrating several innovative features using Flask and the Gemini LLM API. The system includes an Applicant Tracking System (ATS) Scanner, a YouTube Video Summarizer, a Resume Enhancement Tool, a Context-Based Mock Interview System, and an Interactive PDF Chat feature. The ATS Scanner evaluates resumes against job descriptions for compatibility with modern recruitment systems. The YouTube Video Summarizer generates concise notes from video content to facilitate efficient information consumption. The Resume Enhancement Tool offers real-time suggestions to improve resume quality, focusing on formatting and skill highlights. The Context-Based Mock Interview System allows users to practice interviews based on specific job descriptions, enhancing their readiness and confidence. Additionally, the Interactive PDF Chat feature enables users to upload notes in PDF format, allowing the system to train on the content and provide an interactive Q&A experience. This paper details the implementation and integration of these features, highlighting the effectiveness and potential impact on job seekers and professionals in the recruitment proces
Design and Optimization considerations for real-time video conferencing using IMS in 4G/LTE Networks
Get PDFIP Multimedia Subsystem (IMS) offers a plethora of functionalities in 4G/LTE networks. One such functionality is the video conference where three or more User Equipments (UEs) communicate with each other using VoLTE. This research paper presents optimizations in the design and implementation of an IMS video conference server for reducing the video display delay and improving the performance thereof. Both application as well as network level parameters are found to impact the video display delay at the UE, hence performance improvement of both are considered in this paper. Optimization techniques adopted in this implementation resulted in enhanced performance of video conference calls over the LTE network