Exploring the Use of Hypothesis Testing in Determining the Number of Components in Gaussian Mixed Model
Journal Title: Saudi Journal of Engineering and Technology - Year 2017, Vol 2, Issue 12
Abstract
Abstract:Gaussian Mixed Model (GMM) has seen an increase in terms of usage especially to tackle the issue or problem of fraud activities in telecommunication industry. Like any other methods, GMM has its equal share of problems related to the maximum likelihood estimation and the determination of the number of components in GMM. In this paper we will highlight solutions to the said problems such as Expectation Maximization (EM) algorithm and the methods that are normally used to determine the number of components in GMM, which include the most recent research work done by the authors using Kernel method and Akaike Information Criteria (AIC); and the successful derivation of hypothesis testing in the determination of the number of components in GMM. The said derivation has never been attempted before due to the difficulty and complexity of GMM, as exemplified by the use of EM algorithm in solving its maximum likelihood estimation problem. The performance of the hypothesis testing, which is positive and promising despite using different percentage of overlapping; and the comparison of hypothesis testing to AIC, which produced conflicting results under certain conditions will also be highlighted. Keywords:Expectation Maximization (EM), Gaussian Mixed Models (GMM), Kernel method, Akaike Information Criteria (AIC), Hypothesis testing.
Authors and Affiliations
MohdIzhanMohd Yusoff, Ibrahim Mohamed, Abu Bakar
Keywords
Related Articles
- EP ID EP401780
- DOI -
- Views 176
- Downloads 0
Carbon Nanotube and Civil Engineering
Abstract: The construction of roads on soft clay soils is a challenging task and therefore considered as one of the biggest concerns in geotechnical engineering. The utilization of soft soils in tropical regions is curre...
Design and Evaluation of Compressive Strength of Wood Ash Particles (Hardwood) Reinforced Polypropylene (WARPP) Composite Material for Improved Energy Absorption in Crash Energy Management
Abstract:This paper utilized the compressive tests results to examine the crushing behaviour of wood ash particles (hardwood) reinforced polypropylene (WARPP) for crash energy management. The study employed Taguchi metho...
Beyond Hadoop: The Paradigm Shift of Data From Stationary to Streaming Data for Data Analytics
Abstract: The paradigm shift of data from static to fast flowing data is an important move in the industry, to accommodate growing size of data. The velocity and volume of data are continuing to expand which has started...
Survey of Semi Automatic Viscous Fluid Filling Machine
Abstract: Changes in today’s manufacturing environment allow tedious, fatiguing and repetitive tasks to be mechanically performed by robots, as manually controlled work is transition to auto-cycle control equipment. Such...
Study on Liver Data using Clustering Algorithms
Abstract: Data clustering has been considered as the most important raw data analysis method used in data mining technology. To extract the unknown valuable information from the large volume of data for so many real time...