International Journal of Innovations in Science & Technology
Not a member yet
813 research outputs found
Sort by
Predicting Depression Among Type 2 Diabetic Patients Using Federated Learning
Get PDFDepression being a common and dangerous mental health condition could have significant impact on a person\u27s quality of life. It may result in depressive and gloomy feelings along with a loss of interest in once-enjoyable activities. Depression is considered a leading global cause of impairment that affects people at various stages of age, ethnicities, and socioeconomic statuses. It may cause negative effects on person’s physical and emotional well- being like reduced motivation, energy, and appetite. In this paper, we have presented Federated Learning-based framework to predict depression in patients with type 2 diabetes. Type 2 diabetes frequently coexists with depression, which can have a negative impact on treatment outcomes and raise medical expenses. Objective of this paper is to create a Federated Learning- based framework to predict the impact of depression in causing type-II diabetes by analyzing patient’s data that include laboratory results, medical history, and demographic information. To forecast the likelihood of depression in patients with type 2 diabetes. Analysis has been performed using freely available dataset of Type-II diabetes from Kaggle and accuracy of 97% has been achieved
AI-Powered Detection: Implementing Deep Learning for Breast Cancer Prediction
Get PDFBeast cancer remains a critical global health issue, affecting millions of women worldwide. According to the World Health Organization (WHO), there were 2.3 million new cases and 685,000 deaths from breast cancer in 2020 alone. This makes breast cancer the most prevalent cancer globally, with 7.8 million cases diagnosed over the past five years. As the prevalence of breast cancer continues to rise, the need for accurate and efficient diagnostic tools becomes increasingly urgent. Artificial Intelligence (AI) has shown considerable promise in enhancing breast cancer detection and diagnosis. Over the past two decades, AI tools have increasingly aided physicians in interpreting mammograms, offering the potential for automated, precise, and early cancer detection. However, significant challenges remain, particularly concerning data imbalance in datasets—where cancerous images are often underrepresented—and the issue of low pixel resolution, which can obscure crucial details in medical images. This work utilizes a subset of the data called Mini-DDSM, a lightweight version of the Digital Database for Screening Mammography. To address these challenges, our research employed the Neighborhood Cleaning Rule (NCR) algorithm from the imbalance library, designed to mitigate data imbalance by refining the dataset through the selective removal of noisy and borderline examples. This method enhances the quality of training data, enabling AI models to learn more effectively. We developed a deep learning model that incorporates a transfer learning layer (DenseNet121), dense layers, a global pooling layer, and a dropout layer to optimize performance. This model demonstrated promising results, effectively addressing the challenges of data imbalance and low image resolution. Our approach underscores the potential of AI to significantly improve breast cancer detection and diagnosis, ultimately leading to better patient outcomes. Continued research and refinement of AI techniques will be crucial in overcoming remaining challenges and fully realizing the potential of these technologies in healthcare
Internet of Things (IOT) in Developing the Smart Farming and Agricultural Technologies
No full textBackground: The Internet of Things (IoT) is streamlining processes in food and agriculture, especially in developing countries with agriculture-based economies. These countries stand to gain a lot from the IoT innovations that bring about mechanisms to track and control the risks experienced due to factors such as low productivity, wastage of resources, and food scarcity.
Objectives: The purpose of this research paper is to demonstrate how different IoT solutions can be effective in food and agriculture technology in less developed countries. It focuses on the potential of IoT solutions to improve productivity, reduce wastage of resources, and encourage sustainable agro practices. Furthermore, the paper examines the reasons behind the slow adoption of IoT technologies, and strategies to surmount such factors are proposed.
Methodology: The study employed both literature and analytical as well; however, a majority of it was on the primary data concerning IoT solutions that were smart irrigation and precision agriculture. Data was collected from farmers, and technology neglected mostly the politicians of developing countries whose focus was to understand or rather assess the uptake, challenges, and impacts of IoT technology.
Results: The results show that IoT can cause a drastic enhancement of agricultural productivity by efficient water irrigation, keeping a check on soil health and lowering post-harvest waste. MFIs IoT made adoption of the system and use of resources more efficient, increases profits and lowers expenses. However, they also revealed obstacles for the process such as the cost of implementation, expertise in both technical and operational levels and internet services.
Conclusion: Smart agriculture and agricultural systems everywhere will undergo a revolution owing to IoT technologies because it enhances the practices and innovations. However, potential benefits cannot, be maximized Secure fencing of these barriers will not be straightforward since a significant amount of time will have to be devoted to understanding each of the suggestions made by the officers present
Evaluating Faster R-CNN and YOLOv8 for Traffic Object Detection and Class-Based Counting
Get PDFReal-time traffic object detection is a critical component necessary for achieving a fully autonomous traffic system. Traffic object detection, along with background classification, is a significant area of research aimed at enhancing safety on the roads and reducing accidents by accurately identifying vehicles. This research aims to develop an accurate and efficient system for traffic object detection and classification in real-time traffic environments. It also seeks to minimize false positives and negatives, ensuring that no objects are overlooked in the detection of classes such as cars, buses, bicycles, motorcycles, and pedestrians. This research aims and focuses on the two following deep learning technologies: YOLO stands for (You Only Look Once) and Faster R- CNN stands for (Region-based Convolutional neural network). YOLO, initially designed as the single-stage approach, emphasizes speed; therefore, it is best suited for real-time uses. However, Faster R-CNN which is a two-stage detector gives better results in object detection and is highly accurate. Both models are trained and tested on the same data set containing 5712 trained images, 570 validation images, and 270 test images using a workstation with RAM 32 GB and NVIDIA GeForce RTX 4080 Super GPU through the help of CUDA version 12.4 to provide the end evaluating results. Since Faster RCNN is a very intensive model it took 22 hours to complete 3 epochs with an accuracy of 55.2% to train the model and YOLO finished the training within 10 epochs with the [email protected] value of 0.931 of all classes. Our results of traffic object real-time detection indicated that YOLO was vastly better and quicker than Faster R-CNN
Machine Learning-Based Heart Disease Classification for Symptom-Driven Diagnostics
Get PDFHeart diseases are increasing over the period, while identifying cardiac diseases at an early stage continue to pose a challenge. This study focuses on the application of AI specifically in machine learning to improve early diagnosis of this ailment. We overcome limitations of conventional diagnostic paradigms. Normalization was performed on a dataset with demographic and clinical characteristics data, outliers were removed, and principal components analysis was used to enhance and decrease dimensions to get optimized results. The followed classifiers were used: Decision Trees, Random Forests, Logistic Regression, K- Nearest Neighbors, and Naive Bayes, SVM with an assessment of the models based on the confusion matrix, accuracy, and ROC AUC scores. Of all the models created, the Random Forest model was found to have the best internal validation results with an accuracy of 1.0 as well as test and training ROC AUCs of 0.97 for detecting heart disease cases and non-cases. It is evident that developing an AI model for the diagnosis of heart disease provides promising results of faster and efficient diagnosis reducing the mortality rates of the disease
Machine Learning for Detecting Social Media Addiction Patterns: Analyzing User Behavior and Mental Health Data
Get PDFIn the modern world, communication through social networks has become the norm, and people have started to worry about the possible addictive properties of social networks and their influence on mental states. This research aims to propose a Machine Learning (ML) framework for examining patterns of Social Media (SM) addiction, while also acknowledging the dearth of research on developing appropriate detection tools. We obtained data for the research through surveys, which led to the creation of a larger dataset that included aspects of user behavior, mental health parameters, and social media statistics. We use a Random Forest Classifier to predict different levels of addiction, including low, medium, and high levels, while considering behavioral and psychological characteristics. Further analysis of the research findings shows that the more hours spent on social media, especially, are associated with higher levels of distractions, irritation, and other forms of emotional problems among the SM users. Additionally, the feature importance analysis reveals that indicators such as emotional comparisons and the need for self-validation also contribute to addiction. Therefore, these results indicate a high, critical level of awareness and require the development of intervention programs associated with social media addiction while considering the close connection between user behavior and mental health. Lastly, the study adds knowledge on social media addiction and helps to open the next stage in research to identify the prevention of negative impacts on mental health due to addiction to social networks
Analysis of MLP, CNN, and Transfer Learning Using VGG-16 for CIFAR-10 Dataset
Get PDFArtificial Neural Networks (ANN) are becoming the core domain of Artificial Intelligence. Generally, Machine learning and specifically, deep learning gained popularity in problem-solving by virtue of Multi-Layer Perceptron (MLP), Convolutional Neural Networks (CNN), and transfer learning approach. Transfer learning is becoming a powerful and successful technique for a variety of computer vision and image analysis applications due to its capability of reusing well-known proven architectures and their weights. Identification of optimum architecture and classifier along with pre-trained architectures is one of the challenging tasks in achieving optimum accuracy in various image analysis tasks. This paper investigates the performance of MLP, CNN, and transfer learning approaches using VGG-16 by tweaking hyperparameters and classifier architecture. The investigations and critical analysis revealed that MLP and CNN architectures have achieved about 55 % and 80 % validation accuracy on test data. Further experiments using VGG-16 architecture with MLP as a classifier have achieved more than 93 % accuracy on standard specification hardware for image classification on the CIFAR-10 dataset
Codebook-Based Feature Engineering for Human Activity Recognition Using Multimodal Sensory Data
Get PDFRecently, Human Activity Recognition (HAR) using sensory data from various devices has become increasingly vital in fields like healthcare, elderly care, and smart home systems. However, many existing HAR systems face challenges such as high computational demands or the need for large datasets. This paper introduces a codebook-based approach designed to overcome these challenges by offering a more efficient method for HAR with reduced computational costs. Initially, the raw time series data is segmented into smaller subsequences, and codebooks are constructed using the Bag of Features (BOF) approach. Each subsequence is then assigned softly based on the center of each cluster (codeword), resulting in a histogram-based feature vector. These encoded feature vectors are subsequently classified using a Support Vector Machine (SVM). The proposed method was evaluated using the OPPORTUNITY dataset, comprising data from 72 sensors, achieving a classification accuracy of 90.7%. In comparison to other advanced techniques, our approach not only demonstrated superior accuracy in recognizing human activities but also significantly reduced computational costs. The use of soft assignments for mapping codewords to subsequences efficiently captured the key patterns within the activity data. The findings validate that the proposed codebook-based method provides substantial improvements in both accuracy and efficiency for HAR systems
Ransomware Resilience: A Real-time Detection Framework using Kafka and Machine Learning
Get PDFRansomware has emerged as a prominent cyber threat in recent years, targeting numerous businesses. In response to the escalating frequency of attacks, organizations are increasingly seeking effective tools and strategies to mitigate the impact of ransomware incidents. This research addresses the pressing need for real-time detection of ransomware, offering a solution that leverages cutting-edge technologies. The surge in ransomware attacks poses a significant challenge to the cybersecurity landscape, compelling organizations to adopt proactive measures. Recognizing the urgency of the situation, this study motivates the exploration of an innovative approach to ransomware detection. By utilizing advanced tools such as Apache Kafka and Spark, we aim to enhance detection capabilities and contribute to the resilience of businesses against cyber threats. Our methodology employs the Kafka tool and Spark for real-time identification of ransomware exploits. The research utilizes the CIC-MalMem-2022 dataset to develop and validate the proposed model. The integration of Apache Kafka with traditional machine learning techniques is explored to improve the accuracy of cyber threat detection, offering a comprehensive and efficient solution. The implemented model exhibits a commendable detection rate of 95.2%, demonstrating its effectiveness in identifying ransomware attacks in real-time. The combination of Apache Kafka\u27s streaming capabilities and established machine learning methodologies proves to be a potent defense against the evolving landscape of cyber threats. In conclusion, our research provides a robust and practical approach to combating ransomware threats through real-time detection. By leveraging the synergy of Kafka and machine learning, organizations can fortify their cybersecurity defenses and respond proactively to potential ransomware exploits. This study contributes valuable insights and tools to the ongoing efforts in enhancing cyber resilience
Operational Model Based Regional Estimation using Remote Sensing Data
Get PDFWater serves as the vital hub for sustaining life. There is indisputable evidence that the progress of agriculture, which relies directly on water resources, bears direct responsibility for the current global human population. While undeniably invaluable, our planet\u27s freshwater reserves face a mounting challenge in keeping up with the ever-expanding global population. This is primarily due to inefficiencies prevalent in various residential water applications, with irrigation practices in developing nations standing out as a significant contributor to this issue. As our communities continue to grow, it becomes increasingly imperative to address these inefficiencies to ensure sustainable access to this precious resource for generations to come. This dilemma is particularly concerning given the projection of continued population expansion. Concerning irrigation, it is widely acknowledged that more than 60% of water allocated for agricultural purposes is presently being administered in excess, leading to substantial annual wastage. To obtain a precise estimation of the water needed for crop production, it is imperative to devise, develop, and implement a practical and effective method. Employing manual techniques, such as utilizing a lysimeter, for gauging a structure\u27s water requirements is both subjective and financially demanding. This research has been designed to provide a comprehensive measurement of daily ET over a wide geographical area, offering detailed field-specific information. This research work is carried out by utilizing the European Space Agency satellites i.e., Sentinel 2 and 3, and ECMWF meteorological data. The Sentinel-2 data was processed to calculate the biophysical variables, structural parameters, fraction of green vegetation, and aerodynamic roughness. Sentinel 3 data was used to get the land surface temperature. The whole data is then processed to estimate the ET of the chosen area which is discussed in the materials and methods section. Actual water requirement and the water provided to the tobacco crops were compared. The results of the study reveal that estimated ET values were inline with the average surveyed tobacco field values that represents the consistency. However, a significant discrepancy arises due to irregular irrigation practices, indicating a lack of consideration for ET values among farmers. This oversight, coupled with unadjusted irrigation timing and methods, contributes to variance between computed and required ET values, attributed to factors such as human error, insufficient rainfall, and improper practices