Special Issue: VOL.-10, Issue-5, May 2024

Abstract

In this paper, we introduce a managed learning approach for constructing a decision tree aimed at clinical diagnosis. Our primary goal is to develop an efficient classification model with high recall and moderate precision to enhance the efficiency and effectiveness of disease prediction processes. We employed the ID3 algorithm for decision tree construction, and the final model was assessed using standard evaluation methods. This model offers a systematic framework for leveraging relevant information in clinical data, particularly aspects often overlooked by existing methods overly focused on high predictive accuracy. Our analysis was conducted on datasets related to diabetes and coronary disease sourced from the UCI repository. Test results highlight the decision tree's significant contribution to classification quality. Based on these findings, we conclude that decision trees are particularly suitable for addressing disease prediction classification challenges and advocate for their adoption in similar classification tasks.

Keywords: Decision Tree Construction (ID3 Algorithm), Clinical Diagnosis Classification, Managed Learning Approach, Disease Prediction Modeling, UCI Repository, Medical Datasets

References

References not available

Abstract

Mammography stands out as the most cost-effective and efficient method for detecting cancer in its preclinical stages, with breast screening programs specifically designed to identify cancer at earlier stages. These screening programs typically yield vast amounts of data, standardized by the Breast Imaging Reporting and Data System (BI-RADS) established by the American College of Radiology. The BI-RADS system provides a standardized vocabulary for radiologists to use when interpreting each finding. The primary objective of this study is to develop AI models that predict mammography outcomes from a reduced set of interpreted mammography findings. However, the low positive predictive value of breast biopsy results stemming from mammogram interpretation leads to approximately 70% unnecessary biopsies with benign outcomes. In this research paper, data mining classification algorithms, namely Artificial Neural Network (ANN) and Support Vector Machine (SVM), are explored on a mammographic masses dataset. The accuracy of ANN and SVM is reported as 80.3% and 81.9% on test samples, respectively. Our analysis indicates that among these three classification models, SVM predicts the severity of breast cancer with the lowest error rate and highest accuracy.

Keywords: Mammography; Breast Cancer Prediction; Artificial Neural Network (ANN); Support Vector Machine (SVM); BI-RADS Classification.

References

References not available

Abstract

In data mining, one of the main challenges in data preprocessing is handling missing values. Imputation, the process of replacing missing data with substituted values, is crucial for ensuring accurate analysis. Many clinical diagnostic datasets often contain missing values, and excluding these incomplete datasets can introduce more problems than solutions. Traditional imputation methods are easy to implement but may introduce bias in the data. This paper proposes a data imputation technique using K-Nearest Neighbors (KNN) to address the issue of missing data. The method combines KNN predictive modeling with Support Vector Machine (SVM) for improved attribution. The aim of this study is to assess the impact of missing data on the data mining process of learning discovery. Handling missing values in the dataset is a challenging task. Our study explores AI techniques for missing value imputation using Mammogram mass data from the UCI repository. The findings indicate that classifier performance improves when Support Vector Machine (SVM) is employed

Keywords: Missing Value Imputation; K-Nearest Neighbors (KNN); Support Vector Machine (SVM); Mammographic Mass Dataset; Data Preprocessing.

References

References not available

Abstract

Coronary illness, a prevalent global health concern, profoundly impacts human life. With cardiovascular diseases leading to a significant number of deaths worldwide, early and precise diagnosis is crucial for effective prevention and treatment. This study focuses on utilizing the Heart Stalog dataset from the UCI repository and employs Random Forest and Logistic Regression algorithms to predict coronary illness occurrences. Our findings indicate that the Logistic Regression model achieved the highest overall accuracy rate of 83%, outperforming the Random Forest model. This research underscores the importance of developing accurate and efficient classifiers in Data Mining for clinical applications.

Keywords: Coronary illness,Artificial IntelligenceNaive-Bayesian strategies, Neural Networks, Logistic Regression, Support Vector Machines (SVM)

References

References not available

Abstract

This paper introduces a novel feature selection algorithm based on Mutual Information (MI) for both continuous and discrete-valued features. By evaluating the Mutual Information between combinations of features and classes, rather than individual features, the algorithm aims to retain class-discriminative information while reducing the feature set size. The proposed method is applied to the Vehicle dataset from the UCI Machine Learning Repository, where feature dimensionality is reduced using MI-based feature selection followed by a covering technique. The resulting subset of features undergoes preprocessing to ensure data distribution consistency. The classification performance, particularly using Multilayer Perceptron (MLP), demonstrates promising results compared to conventional classifiers and approaches using individual feature information.

Keywords: Mutual Information (MI) Feature Selection Multiclass Classification Multilayer Perceptron (MLP) Clinical Data Integration

References

References not available

Abstract

Cancer is a leading cause of mortality globally, with breast cancer posing a significant threat to women's health worldwide. Early detection is key to effective treatment. This study employs machine learning techniques, specifically Multilayer Perceptron (MLP) and Support Vector Machine (SVM) classifiers, to classify breast cancer data. Utilizing SVMRFE for dimensionality reduction, the study aims to identify the smallest subset of features for improved classification of benign and malignant tumors. By analyzing the Wisconsin Breast Cancer (WBC) dataset, this research seeks to optimize feature selection methods to enhance classification accuracy. Results indicate that MLP classifier achieves higher accuracy rates post feature selection. Comparative analysis of SVM and Artificial Neural Network underscores the efficacy of feature selection techniques in improving classification performance.

Keywords: Breast Cancer Detection Machine Learning Support Vector Machine (SVM) Multilayer Perceptron (MLP) Feature Selection (SVM-RFE)

References

References not available

Abstract

Mammography, a key tool in early threat detection, drives chest screening programs aimed at spotting infections early. These programs, managed by BI-RADS, yield vast data analyzed by radiologists. This study focuses on AI models predicting mammography outcomes, aiming to reduce unnecessary biopsies. Naïve Bayes and K-Nearest Neighbor algorithms were tested, with Naïve Bayes showing the highest accuracy at 85.43%.

Keywords: Mammographic Mass Analysis Breast Cancer Screening BI-RADS Classification Naïve Bayes Classifier K-Nearest Neighbor (KNN)

References

References not available

Abstract

This paper introduces a supervised learning approach for constructing decision trees in clinical diagnosis. The primary aim is to develop an efficient classification model with a balance of high recall and moderate precision to enhance the effectiveness of disease prediction. Utilizing the ID3 algorithm for decision tree construction, the final model is evaluated using standard assessment methods. This model offers valuable insights into leveraging clinical data, particularly aspects often overlooked by existing techniques focused solely on high precision. Experiments conducted on diabetes and coronary heart disease datasets from the UCI repository demonstrate the decision tree's effectiveness in classification tasks. Based on these findings, we conclude that decision trees are well-suited for addressing disease prediction challenges and recommend their adoption in similar classification problems.

Keywords: Supervised Learning Decision Tree (ID3 Algorithm) Clinical Diagnosis Disease Prediction Classification Performance (Recall and Precision)

References

References not available

Abstract

Breast cancer ranks as the most prevalent cancer type among women worldwide, being the second highest cause of female mortality among all cancer types. Accurately predicting the survival rate of breast cancer patients is a critical concern for cancer researchers. Machine Learning (ML) has garnered considerable attention for its potential to provide precise results, yet its methodologies and predictive performance remain debatable. This paper focuses on employing ML algorithms to predict Haberman's Breast Cancer Survival study. Specifically, two different ML approaches, namely Multilayer Perceptron (MLP) and Support Vector Machine (SVM) models, are explored for Breast Cancer Survival prognosis. The classification performance of abnormal and normal Breast Cancer Survival patients is assessed across various metrics including training and testing accuracy, precision, and recall. The objective of this systematic review is to identify and critically evaluate current studies regarding the application of ML in predicting the 5-year survival rate of breast cancer. Test results on Haberman's Breast Cancer Survival dataset demonstrate the effectiveness of the proposed MLP approach, achieving an accuracy of 97.54%.

Keywords: Prediction, SVM, MLP and ML

References

References not available

Abstract

The effectiveness of AI models heavily relies on the selection of key features within the dataset. Feature selection is pivotal in model optimization, aiming to identify a concise set of features to build robust models with minimal redundancy. Utilizing advanced algorithms for feature selection enhances the predictive speed of the models, mitigating the impact of redundant and noisy data that can hinder data understanding and model efficacy. Consequently, practitioners seek to extract meaningful insights from vast datasets using feature selection techniques. Feature selection involves identifying the most relevant features while eliminating redundant and irrelevant ones. In this study, we conducted a comparison between SVMRFE (Recursive Feature Elimination) based feature selection method using a prominent dataset (Liver disease and hepatitis dataset). Two classification algorithms, decision tree, and Naive Bayes, were employed to evaluate the performance of the algorithms. Results indicated that the decision tree classifier achieved higher accuracy rates on the dataset following the application of feature selection methods. The analysis highlights the efficacy of feature selection techniques in enhancing the performance of learning algorithms.

Keywords: Feature Selection SVM–Recursive Feature Elimination (SVM-RFE) Liver Disease Prediction Decision Tree Classifier Naïve Bayes Algorithm

References

References not available

Abstract

Coronary disease stands as one of the most significant human health challenges worldwide, profoundly impacting human lives. Cardiovascular disorders, including heart-related ailments, have been responsible for a considerable number of deaths globally over the past few decades, emerging as the deadliest disease not only in India but also worldwide. Accurate and timely diagnosis of coronary disease is crucial for preventing cardiovascular failure and ensuring effective treatment. Thus, there is a pressing need for robust, precise, and efficient systems to diagnose such infections promptly for appropriate treatment. In this study, we utilized the Heart Stalog dataset obtained from the UCI repository, employing Neural Networks and Logistic Regression algorithms to accurately predict the occurrence of coronary disease. The proposed decision support system based on Neural Networks and Logistic Regression will aid healthcare professionals in efficiently identifying heart patients. Logistic Regression emerged as the superior model among the two algorithms, achieving an overall accuracy rate of 91.54%. Our findings demonstrate the superior performance of Logistic Regression over Neural Networks in terms of precision. Developing accurate and computationally efficient classifiers for clinical applications remains a significant challenge in Machine Learning.

Keywords: Coronary Disease Prediction Artificial Intelligence in Healthcare Logistic Regression Artificial Neural Networks (ANN) Heart Disease Classification

References

References not available

Abstract

Medical image fusion enhances image reliability by integrating key data from multiple sources. This paper introduces a fusion method using NSST (Non-Sampled Shearlet Transform) in the transform domain. Entropy analysis, based on PCNN (Pulse Coupled Neural Network), is employed to optimize fusion. The proposed algorithm prioritizes highinformation bands for fusion, resulting in superior image quality. NSST and PCNN are applied to fuse MRI and PET images, ensuring comprehensive data integration.

Keywords: Medical Image Fusion Positron Emission Tomography (PET) Magnetic Resonance Imaging (MRI) Non-Subsampled Shearlet Transform (NSST) Pulse Coupled Neural Network (PCNN)

References

References not available

Abstract

This research focuses on early detection of heart disease symptoms using patient data and real-time user input. Modern healthcare data includes detailed demographic and symptom information, allowing for comprehensive analysis. Our proposed method utilizes this data for classification, comparing recent healthcare data with baseline distributions. Machine learning techniques such as Logistic Regression, K-Nearest Neighbors, Random Forest, and XGBoost are employed for training and testing. Classifier performance is evaluated to make predictions.

Keywords: Heart Disease Identification E-Healthcare Systems Machine Learning Classification Random Forest and XGBoost Logistic Regression and K-Nearest Neighbors (KNN)

References

References not available

Abstract

Brain tumors pose significant health risks, requiring effective treatment planning for patient survival. Medical imaging techniques like MRI assist in tumor diagnosis, but manual classification is hindered by data volume. Automatic classification schemes, such as CNN, are crucial for accurate detection and treatment.

Keywords: Brain Tumor Detection Magnetic Resonance Imaging (MRI) Convolutional Neural Networks (CNN) Medical Image Classification Machine Learning in Healthcare

References

References not available

Abstract

This study proposes a new method for Alzheimer’s Disease detection using 3D brain MR images and first-order statistical features. Alzheimer’s is a progressive neurodegenerative disorder affecting the elderly. The method focuses on extracting features from grey and white matter to predict AD using ensemble classifiers.

Keywords: Alzheimer’s Disease Detection 3D Brain MRI Image Processing Techniques First-Order Statistical Features Ensemble Classifiers

References

References not available

Abstract

LBW, indicating newborn health issues, is linked to infant mortality and long-term health concerns. This study utilizes machine learning to detect potential LBW cases early by analyzing maternal health indicators. It frames the problem as a binary classification task and achieves improved accuracy. Decision rules derived from Indian healthcare data aid in predictive healthcare for smart cities, with a screening tool developed for Obstetrics and Gynecology professionals.

Keywords: Low Birth weight (LBW), Smart health informatics, Predictive analytics, Machine Learning (ML).

References

References not available

Abstract

Diabetes mellitus is a chronic metabolic disorder that affects millions of people worldwide, posing significant health challenges and economic burdens. Early detection and accurate prediction of diabetes are crucial for effective management and prevention of complications. Machine learning techniques offer promising solutions for predicting diabetes risk based on patient data. In this research paper, we present a comprehensive analysis of machine learning algorithms for predicting diabetes mellitus using the Pima Indians Diabetes Database available on Kaggle. The dataset comprises various biomedical attributes such as glucose concentration, blood pressure, and insulin levels, collected from Pima Indian women. Five machine learning algorithms, including Logistic Regression, Support Vector Machines, Random Forest, K-Nearest Neighbors, and Gradient Boosting, are implemented using Python. The performance of each algorithm is evaluated using multiple metrics, and the results are analyzed to identify the most effective model for diabetes prediction. This study provides valuable insights for healthcare professionals and researchers in the field of diabetes management and predictive analytics.

Keywords: Diabetes Mellitus Prediction Pima Indians Diabetes Database Machine Learning Algorithms Comparative Classification Analysis Predictive Healthcare Analytics

References

References not available