Comparing the Accuracy of Hierarchical Agglomerative and K-means Clustering on Mobile AR Usability Metrics
Abstract
This article presents the experimental work of comparing the performances of two machine learning approaches, namely Hierarchical Agglomerative Clustering and K-means Clustering on Mobile Augmented Reality Usability datasets. This research will present the background, related literature, initial findings, objectives and the proposed methodology, before presenting the evidences and discussion of the intended research.
Key Words and Phrases: Machine Learning, Hierarchical Agglomerative Clustering, K-means, Usability, Performance Metrics, Self-reported Metrics
Improved Feature Extraction Method for Classification of Individual Activity Recognition
Abstract
The application of context-aware and smartphones using embedded sensors have the potentials to monitor tasks that are practically complicated to access. Human beings may not be safe except they are informed early of impending dangers around them. However, inaccuracy of individual activity recognition (IAR) from accelerometer sensor signals data and hidden information present in the data due to classification problem still limits the opportunities for efficient prediction of individual abnormality behaviour in a crowd scenario. This paper proposes an improved feature extraction method with statistical-based time frequency domain (SBTFD) that represent and extract hidden contexts information from sensor signals with improved accuracy for classification of individual activities in a crowd-prone area based on activity recognition (AR).The experiment on the proposed method of the SBTFD with J48 achieved a maximum of 99:2% accuracy outperforms the state-of- the-art method in activity recognition.
KeyWords: Individual activity recognition (IAR); Statistical based time frequency domain (SBTFD); Context-aware; Smartphone and Crowd scenario.