Hashibah Hamid

Principal Component Analysis – A Realization of Classification Success in Multi Sensor Data Fusion

The field of measurement technology in the sensors domain is rapidly changing due to the availability of statistical tools to handle many variables simultaneously. The phenomenon has led to a change in the approach of generating dataset from sensors. Nowadays, multiple sensors, or more specifically multi sensor data fusion (MSDF) are more favourable than a single sensor due to significant advantages over single source data and has better presentation of real cases. MSDF is an evolving technique related to the problem for combining data systematically from one or multiple (and possibly diverse) sensors in order to make inferences about a physical event, activity or situation. Mitchell (2007) defined MSDF as the theory, techniques, and tools which are used for combining sensor data, or data derived from sensory data into a common representational format. The definition also includes multiple measurements produced at different time instants by a single sensor as described by (Smith & Erickson, 1991).

New Location Model Based on Automatic Trimming and Smoothing Approaches

Location Model is a classification approach that capable to deal with mixed binary and continuous variables at once.The binary variables create segmentation in the groups called cells whilst the continuous variables measure the differences between groups based on information inside the cells.It is important to note that location model is biased and even impossible to be constructed when involving some empty cells.Interestingly from previous studies, smoothing approach managed to remedy the effects of some empty cells.However, numerical analysis has demonstrated that the performances of the location model based on smoothing approach are good in most situations except if there are outliers in the sample.Thus, the presence of outliers has alarmed us to further investigating the performance of the location model.Therefore, in this paper, we develop a new methodology of location model producing new model called automatic trimmed location model through new estimators resulting from an integration of automatic trimming and smoothing approaches in addressing both issues of outliers and empty cells simultaneously.The results have confirmed that the new methodology developed as well as the new location model produced offer another potential tools to practitioners, which possible to be considered in classification problems when the data samples contain outliers and at the same time could resolve the crisis of some empty cells of the location model. Copyright

Winsorized and Smoothed Estimation of the Location Model in Mixed Variables Discrimination

View references (45)The location model is a familiar basis and excellent tool for discriminant analysis of mixtures of categorical and continuous variables compared to other existing discrimination methods.However, the presence of outliers affects the estimation of population parameters, hence causing the inability of the location model to provide accurate statistical model and interpretation as well.In this paper, we construct a new location model through the integration of Winsorization and smoothing approach taking into account mixed variables in the presence of outliers.The newly constructed model successfully enhanced the model performance compared to the earlier developed location models. The results of analysis proved that this new location model can be used as an alternative method for discrimination tasks as for academicians and practitioners in future applications, especially when they encountered outliers problem and had some empty cells in the data sample.

New algorithm of Location Model based on robust estimators and smoothing approach

Location Model is a classification model that capable to deal with mixtures of binary and continuous variables simultaneously. The binary variables create segmentation in the groups called cells whilst the continuous variables measure the differences between groups based on information inside the cells. It is important to note that location model is biased and even impossible to be constructed when involving some empty cells. Interestingly from previous studies, smoothing approach managed to remedy the effects of some empty cells. However, numerical analysis has demonstrated that the performances of the location model based on smoothing approach are good in most situations except if there are outliers in the sample. Thus, the presence of outliers has alarmed us to do further investigation towards the performance of the location model. Instead of transformations or truncation, many researchers used various robust procedures to protect their data from being distorted by outliers. Therefore, in this paper, we develop a new methodology of the location model through new estimators resulting from an integration of robust estimators and smoothing approach to address both issues of outliers and empty cells simultaneously. It is expected that this new methodology will offer another potential tool to practitioners, which is possible to be considered in classification problems when the data samples contain outliers and at the same time could resolve the crisis of some empty cells of the location model.

A framework of mixed variables classification in the presence of outliers: A robust location model

The natural outstanding of location model is as an excellent tool for mixed variables classification among other existing approaches such as Kernel-based non-parametric classification, logistic discrimination and linear discriminant analysis. However, the presence of outliers will affect the estimation of population parameters, hence causing inability of the model to provide an adequate statistical model and interpretation as well. In other words, outliers can distort not only parameters estimation, but also lead to poor classification performance. Therefore, this article aims to develop a new framework of location model through the integration of robust technique and classical location model with mixed variables in the presence of outliers, purposely for robust classification. The developed framework produces a new location model that can be used as an alternative approach for classification tasks as for academicians and practitioners in future applications, especially when they are facing with outliers ...

A New Framework of Smoothed Location Model with Multiple Correspondence Analysis

The implication of a considering large binary variables into the smoothed location model will create too many multinomial cells or lead to high multinomial cells and more worrying is that it will cause most of them are empty. We refer this situation as large sparsity problem. When large sparsity of multinomial cells occurs, the smoothed estimators of location model will be greatly biased, hence creating frustrating performance. At worst, the classification rules cannot be constructed. This issue has attracted this paper to further investigate and propose a new approach of the smoothed location model when facing with large sparsity problem.

The performance of smoothed location model with PCA+Indicator MCA and PCA+Adjusted MCA

: Smoothed location model (SLM) is one of the discriminant analysis that can be used to deal with mixtures of continuous and binary variables simultaneously. However, SLM facing the problem in estimating parameters when the there is a large number of binary variables considered in the study. Thus, two variable extraction techniques, principal component analysis (PCA) and multiple correspondence analysis (MCA) are conducted together with SLM in order to solve the problems of many empty cells and parameters estimation. Simulation results showed that SLM along with PCA+Adjusted MCA performed better than SLM with PCA+ Indicator MCA even when the number of extracted binary is large.

Variables extraction on large binary variables in discriminant analysis based on mixed variables location model

The natural performance of the location model is a potential tool for allocating an object into one of the two observed groups involving mixtures of continuous and binary variables. In constructing location model, continuous variable is used to estimate parameters while binary variable is utilized to create segmentation in each group. Such segmentation is called as multinomial cells. Basically, the multinomial cells will grow exponentially according to the number of the binary variable. These multinomial cells will become empty when there is no object can be assigned into some of them. Then the occurring of empty cells will lead to unreliable parameter estimation. Consequently, the construction of the discriminant rule based on location model is impossible. Therefore, this paper attempts to discuss how the location model based on maximum likelihood estimation can be constructed even dealing with many measured binary variables. In other word, how is location model able to deal with the issue of many empty ...

Multiple correspondence analysis for handling large binary variables in smoothed location model

Smoothed location model is a discriminant analysis which can be used to handle the data involving mixtures of continuous and binary variables simultaneously. This model is introduced to handle the problem of some empty cells due to the increasing of binary variables. However, smoothed location model is infeasible if involve large number of binary variables. Therefore, the combination of two variable extraction approaches, principal component analysis and multiple correspondence analysis are carried out before the construction of smoothed location model in order to extract large number of measured variables in the study. In fact, there are four types of multiple correspondence analysis but only Burt matrix multiple correspondence analysis had been applied in the latest investigation. Thus, this study aims to examine and compare principal component analysis with four types of multiple correspondence analysis and hope to have better results for data with large number of mixed variables. The proposed model is...