Browsing by Author "Faizullah Farhan"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Embargo
    Fall Prediction Using Machine Learning Method
    (North South University, 2023) Faizullah Farhan; Jubaer Al Noman; DR. MOHAMMAD MONIRUJJAMAN KHAN; 1931708642; 1931536642
    The occurrence of falls among the elderly population poses a substantial concern in the realm of public health due to the high prevalence of severe injuries and the subsequent negative impact on their overall well-being and quality of life. The significance of establishing procedures to reduce falls becomes more pronounced with the rise in the older population. This research study comprehensively examines the application of multiple machine-learning algorithms for fall detection and prevention. In this study, we are looking into nine popular machine-learning techniques. These include support vector machines (SVM), random forests, naive Bayes, logistic regression, and linear discriminant analysis (LDA), voting classifiers, K-nearest neighbors (KNN), AdaBoost, and gradient boosting. The implemented algorithms were trained using a dataset that included factors relevant to fall detection, such as acceleration, direction, and positional data. The experiments included a dataset of samples from wearable devices equipped with sensors, representing both fall and non-fall scenarios in real-world settings. A complete evaluation methodology was employed, incorporating cross-validation techniques and performance measures such as accuracy to ensure the correctness and reliability of our findings. The results obtained from our study have revealed the promising capabilities of machine learning in the context of fall prevention systems, yielding positive outcomes. The Random Forest, Gradient Boosting, and Voting Classifier models exhibited the best accuracy rates, with a 97% accuracy in fall detection. Support vector machines, logistic regression, LDA, and AdaBoost performed exceptionally, achieving accuracy levels ranging from 95% to 97%. In contrast to other approaches, Naive Bayes had a comparatively lower accuracy rate of 48%. On the other hand, our study achieved 98% accuracy using a deep CNN model. Based on our research findings, it is commonly observed that ensemble techniques, such as the random forest model, the gradient boosting model, and the CNN model, tend to exhibit superior performance compared to individual algorithms. Moreover, the findings underscore the importance of feature engineering and choosing appropriate machine learning algorithms to augment fall prevention systems' efficacy. This research contributes to the existing knowledge of fall prevention by providing intuition into the suitability of different machine-learning methods for real-time fall detection and prevention. The discovery of this study 8 can provide valuable insights for researchers, practitioners, and healthcare professionals in their selection of optimal algorithms for the effective implementation of dependable fall prevention systems. The present work undertook a detailed assessment of various classification models within the domain of fall prevention, employing machine learning methodologies. The Decision Tree and Random Forest models demonstrated high levels of accuracy, with rates of 94% and 96%, respectively. This highlights their robust predictive capabilities in the domain of fall detection. In addition, Logistic Regression, Gradient Boosting, AdaBoost, Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), Gaussian Process, Support Vector Machines (SVM), K Nearest Neighbors (KNN), and Multilayer Perceptron (MLP) models exhibited recall, significant precision, and F1-score metrics, showing their embryonic suitability in diverse fall detection scenarios. Despite Naive Bayes demonstrating relatively lower accuracy, it displayed noteworthy recall values, suggesting its potential suitability in specific fall detection scenarios. The findings above offer valuable insights for healthcare professionals and researchers aiming to apply machine learning techniques to prevent falls.