International Journal of Innovations in Science & Technology
Not a member yet
813 research outputs found
Sort by
Credit Card Fraud Detection Using Machine Learning with Undersampling and SMOTE Oversampling
Get PDFCredit card fraud detection is currently the most popular implementation domain of Computational Intelligence techniques. A common issue in the present world is being faced by many organizations and institutions. This is due to the increase in the frequency of transactions, which are now conducted electronically and a higher increase in the number of electronic commerce platforms. In the present world, we are experiencing many credit card issues. In this paper, we apply various algorithms of machine learning as random forest, logistic regression and k-Nearest Neighbors (KNN) to train the specified machine learning model using a given dataset to design the comparative conducted on the accuracy and various measures of the models as it is being implemented via each of such algorithms. To address this, we evaluate the possibility of under-sampling and SMOTE as approaches that can enhance multiple machine-learning models. An accuracy of 99.99% in the dataset was achieved using the SMOTE technique with the Random Forest model. This research concludes that SMOTE improves the performance of the machine learning model for fraud identification and presents a more efficient approach to address the problem of class imbalance
Exploring Deep Learning Approaches for Early Detection of Chronic Kidney Disease: Trends and Techniques
No full textThis study investigates the application of deep learning models, namely CNN, RNNs, and MLP, for the early prediction of CKD. Early detection of CKD is critical for initiating timely treatment, as the disease can advance with few symptoms. The research leverages a preprocessed Kaggle dataset, divided for training and testing, to assess model performance. Among the models, CNN achieved an impressive 99% accuracy, highlighting its strong feature extraction capabilities. The RNN and MLP models also demonstrated high accuracy, reinforcing the potential of deep learning in enhancing CKD screening processes. This approach can support more personalized and preventive healthcare, potentially improving patient outcomes through earlier interventions.
Keywords: RNN, CKD, Deep Learning, CNN, ANN, LSTM, Performance Optimization
Abbreviation
Full Form
CKD
Chronic Kidney Disease
CNN
Convolutional Neural Network
RNN
Recurrent Neural Network
MLP
Multi-Layer Perceptron
ANN
Artificial Neural Network
LSTM
Long Short-Term Memory (a type of RNN
Artificial Intelligence-Based Approach for The Recommendations of Mango Supply Chain
Get PDFThis study utilizes a comprehensive dataset that encompasses variables reflecting temperature, humidity, precipitation, inventory levels, transportation modes, freshness scores, and ripeness scores. Compiled from various mango farms across different markets, this dataset provides a robust foundation for our analysis. To develop predictive models, we employed several machine learning algorithms, including Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Random Forests (RF), and Decision Trees (DT). We divided the dataset into training and testing sets, using an 80-20 split for training and testing subsets, respectively. Model performance was evaluated using metrics such as accuracy, precision, recall, and F1 score. Our results indicate that Random Forests outperformed other models, achieving the highest accuracy, precision, recall, and F1 scores. A feature importance analysis revealed specific features that contributed significantly to the performance improvements of the model. These insights into feature importance can aid in refining the model\u27s performance, making feature importance analysis a valuable component of model evaluation
Austempering Time and Its Influence on the Mechanical Performance of Inverse Bainite: Insights from Hardness, Toughness, and Strength Testing
Get PDFThis study examines the impact of austempering time on the mechanical properties of 0.8C experimental steel with inverse bainitic microstructures. Samples were austenitized at 900 °C and austempered at 420 °C for 30, 60, 90, and 120 minutes. The effects on hardness, impact toughness, and (UTS) were analyzed. Microstructural evaluations by optical microscope, scanning electron microscope, and energy dispersive spectroscopy confirmed the formation of inverse bainite. Mechanical testing showed that increasing austempering time leads to higher hardness due to cementite lath growth and reduced ferrite content, but also results in lower Charpy impact energy, indicating reduced toughness. While UTS initially decreases, it increases sharply after 120 minutes, accompanied by brittle fracture. This study suggests that prolonged austempering enhances hardness and strength but increases brittleness, making these structures less suitable for applications involving sudden forces
Prediction of Brain Stroke Using Federated Learning
Get PDFStroke often arises from an abrupt blockage in the blood vessels supplying the brain and heart. Detecting early warning signs of stroke can significantly reduce its impact. In this study, we propose an early prediction method for stroke using various Machine Learning (ML) techniques, considering factors such as hypertension, body mass index, heart disease, average glucose levels, smoking habits, prior stroke history, and age. These attributes, rich in information, were utilized to train three distinct classifiers: Logistic Regression, Decision Tree, and K-nearest neighbors for stroke prediction. In this study, Federated Learning (FL) has been applied to combine the ML models (Logistic Regression (LR), Decision Tree (DT), and K-Nearest Neighbors (KNN)) from distributed medical data sources while preserving patient privacy. By aggregating locally trained models from multiple hospitals or devices, FL ensures the robustness of the weighted voting classifier without requiring direct data sharing, thereby enhancing stroke prediction accuracy across diverse datasets. Subsequently, the results from these base classifiers were combined using a weighted voting approach to achieve the highest accuracy. Our study demonstrated an impressive accuracy rate of 97%, with the weighted voting classifier outperforming the individual base classifiers. This model proved to be the more accurate in predicting strokes. Additionally, the Area Under the Curve (AUC) value for the weighted voting classifier was notably high, and it exhibited the lowest false positive and false negative rates compared to other classifiers. Consequently, the weighted voting classifier emerged as an almost ideal tool for predicting strokes, offering valuable support to both physicians and patients in identifying and preventing potential stroke incidents
AI-Driven Control and Processing System for Smart Homes with Solar Energy
Get PDFIn recent years, the utilization of solar energy has grabbed attention in the industrial and domestic zones. The existing systems to use the services of solar cells are conventional. These systems require parameters (irradiance and temperature) for desirable results that are unknown to the end user. These parameters change with regions and human to human. Therefore, an Artificially Intelligent, Control and Processing System is designed to get more accurate results with the unique feature of empowering the end user, which uses the parameters assembled on different regions. The proposed system has an improved PV model based on (ANN) that resembles experimental results with a few readily available, reprogrammable input parameters from the PV module datasheet. The developed system uses regional irradiation data which exhibits minimal fluctuations. In the model presented here; to avoid overburdening problems, loads were divided into manageable chunks MK. In this case load chunks (needed) were moved from solar to utility more stably and economically. Briefly stated the suggested solution provides a complete package for integrating solar energy systems with the grid in an automated and resilient way
Enhancing Mobile Efficiency: A Cloud-Powered Paradigm for Extended Battery Life and Enhanced Processing Capabilities
Get PDFIn an interconnected world where mobile phones are essential to everyday operations, the constraints of these devices in terms of processing power, memory, storage, and energy efficiency are becoming increasingly apparent. This research introduces an innovative solution by integrating Mobile Cloud Computing (MCC) to address these challenges. The research focuses on the creation of an Android application called "ServiVerse" that efficiently drains the phone\u27s battery to imitate real-world conditions. The software is accompanied by a Firebase-connected battery optimizer, which provides users with complete insights into battery state, cleaning history, and graphical representations of performance. The system\u27s distinguishing feature is outsourcing power-intensive operations to a cloud server, resulting in increased energy efficiency and battery life. The study demonstrated successful battery optimization tactics adapted to individual users, such as the amount of cache and RAM deleted and storage space freed up on the mobile devices. This strategy has proven to be vital in addressing a key concern about background processing and the loss of power generation on mobiles, which is providing users with more efficient and longer-lasting battery life
Bitcoin Price Forecasting: A Comparative Study of Machine Learning, Statistical and Deep Learning Models
Get PDFIntroduction/Importance of Study:
Cryptocurrency price prediction is crucial for investors and researchers, given the market\u27s nonlinear nature and the potential for significant financial implications.
Novelty:
This study offers a novel approach to cryptocurrency price prediction, leveraging a range of machine learning and deep learning models to address the challenges of predicting Bitcoin\u27s exchange rate.
Materials & Methods:
The study employs various machine learning and deep learning models, including Recurrent Neural Networks (RNNs) such as Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), along with traditional models like Linear Regressor, Random Regressor, ExtraTreesClassifier, XGBoost Regressor, ARIMA, Prophet, and CNN.
Results & Discussion:
The ExtraTreesClassifier model emerged as the top performer, achieving a Test MAPE of 0.0689. This model outperformed deep learning models like RNNs, indicating its effectiveness in cryptocurrency price prediction.
Conclusion:
The findings suggest that the proposed models, particularly the ExtraTreesClassifier, can provide valuable insights for investors and traders in the cryptocurrency market
A Deep Learning Based Mobile Application for Wheat Disease Diagnosis
Get PDFWheat is one of the major staple crops in Pakistan, playing a crucial role in ensuring food security and contributing to the country\u27s economy. The productivity and quality of wheat crops, however, are vulnerable to several illnesses. The ability to diagnose these diseases quickly and accurately is crucial for taking the appropriate preventative actions, limiting losses, and maintaining food security. In this research paper, we build and test a wheat disease detection system adapted to the conditions in Pakistan. The suggested method uses machine learning-based techniques along with image processing algorithms to automatically detect and categorize various wheat diseases based on their symptoms. High-resolution photos of healthy wheat plants and sick plants displaying different diseases were collected from different regions of Pakistan in order to construct an accurate and robust disease detection model. The dataset has been annotated by plant pathologists who provided true labels for use in evaluation and training. To achieve the best results in wheat disease diagnosis, many cutting-edge deep-learning architectures were investigated and optimized. These included Convolutional Neural Networks (CNNs) and Transfer Learning models. Multiple models’ effectiveness was evaluated using accuracy, precision, and recall, in a series of extensive trials
Breast Masses Detection Using YOLOv8
Get PDFBreast cancer stands as a formidable global health challenge, necessitating swift and precise diagnostic measures to combat its devastating impact. In this study, we delve into the efficacy of YOLOv8, a cutting-edge artificial intelligence model, for the precise detection and localizing of breast masses in digital mammography images. YOLOv8’s inherent capability to simultaneously detect and localize masses showcases accurate pinpointing of the exact locations of abnormalities within mammographic scans. Our comprehensive evaluation reveals compelling performance metrics, including an F1 score of 0.91 and a mean Average Precision (mAP) of 0.942. These results depict the robustness of the YOLOv8 in mass detection but also show better results than the conventional clinical methods, offering higher accuracy and efficiency in the diagnostic process. This study explains the transformative potential of YOLOv8 in revolutionizing breast cancer detection paradigms, presenting a promising pathway toward enhancing early detection rates and ultimately improving patient outcomes.
abnormalities within mammographic scans. Our comprehensive evaluation reveals compelling performance metrics,
including an F1 score of 0.91 and a mean Average Precision (mAP) of 0.942. These results depict the robustness of the
YOLOv8 in mass detection but also show better results than the conventional clinical methods, offering higher accuracy
and efficiency in the diagnostic process.
This study explains the transformative potential of YOLOv8 in revolutionizing breast cancer detection paradigms, presenting a promising pathway toward enhancing early detection rates and ultimately improving patient outcomes