No abstract

Kalman filter is normally used to enhance speech quality in a noisy environment, in which the speech signals are usually modelled as autoregressive (AR) process, and represented in the state-space domain. It is a known fact that to identify the changing AR coefficients in every time state requires extensive computation. In this paper, the authors develop a bidirectional Kalman filter and apply it in a speech processing system. The proposed filter uses a system dynamics model that utilises the past and the future measurements to form an estimate of the system's current time state. It provides efficient recursive means to estimate the state of a process that minimises the mean of the squared error. Compared to the conventional Kalman filter, the proposed filter reduces the computation time in two ways: (i) by avoiding the computation of AR parameters in each time state, and (ii) by reducing the dimension of the matrices involved in the difference equations and the measurement equations into constant (1 × 1) matrices. The speech recognition result shows that the developed speech recognition system becomes more robust after the proposed filtering process, and the proposed filter's low computational expense makes it applicable in the practical hidden Markov model-based speech recognition system.

Schooling system must provide high quality learning opportunities to meet the educational needs and ensuring achievement for every student. All teachers monitor their students’ progress throughout the year, includes formative assessment, questioning, providing feedback, etc. This practice helps teachers continually assess students’ academic performance and evaluate the effectiveness of their teaching. In this paper, k-means clustering method with deterministic model is used to analyze the student's overall performance. The results is important for educators to identify students who are at risk academically and areas where teaching strategies may need adjustment to better meet these students' needs.

No abstract

Schooling systems always offer finest teaching and learning opportunities to reach the educational requirements and ensuring achievement for every student. However, health affects students' academic performance directly. All teachers monitor their students' progress throughout the year, includes formative assessment, attendance rates, involvement in the organization, etc. This practice helps teachers continually assess the conditions of students and their academic performance. Data mining is a process to explore certain style and hidden correlation among massive volume of data. Data mining is applied in multiple disciplinary fields, for example, insurance, education, banking and bioinformatics. Data mining skills such as clustering, classification, regression and prediction are commonly used by educators to measure academic performance. In this paper, method of k-means clustering with deterministic model is applied to analyze the student's overall performance. The students' assessment scores are assigned to k clusters without prior knowledge of the scores. The result is important for educators to further investigate the effect of sickness of students within a cluster that may lead to poor academic performance.

Regression analysis is a statistical methodology to investigate the relationship between the dependent variable and the independent variables. In current era with the trend of big data, we might face some problems when performing statistical analysis for the massive volume of data. For example, the heavy burden of the computing load will cause the computation to be time consuming, the accuracy of the results might be affected in view of the vast volume of data. Hence, divided regression analysis is proposed to reduce the burden of the computing load. This approach performs subdivision of the dataset into several unique subsets, then the multiple linear regression is fi tted into each subset. The results obtained from each subset are then combined to obtain a divided regression model which is treated as the original overall dataset. The dataset used in this paper is KC Housesales Data, obtained from the Kaggle website. The dataset contains statistics information about the housing price, for example, size of lot, size of living area and selling price of the house. The goal of this paper is to predict the selling price of a house from the given attributes. The dataset is partitioned into fi ve subsets. Consequently, multiple linear regression is fi tted for each subset. Then, some model adequacy checking will be applied on the models. The test in determining the existence of multicollinearity in the models is rather important as well because the collinearity among the independent variables will affect the overall results. Hence, the variance infl ation factor (VIF) approach is used to determine the existence of multicollinearity. Finally, the divided regression model is obtained by combining results from all the subsets and the validity of divided regression model is verifi ed.

The f -divergence of Csiszar is defined for a non-negative convex function on the positive axis. A pseudo f -divergence can be defined for a convex function not satisfying the usual requirements. A rational function where both the numerator and the denominator are non-integer polynomials will be used to generate universal portfolios. Five stock-price data sets from the local stock exchange are selected for the empirical study. Empirical results are obtained by running the generated portfolios on these data sets. The empirical results demonstrate that it is possible for the investors to increase their wealth by using the portfolios in investment.

Confidence intervals for the γ-quantile of a linear combination of N non-normal variates with a linear dependence structure would be useful to the financial institutions as the intervals enable the accuracy of the value at risk (VaR) of a portfolio of investments to be quantified. Here we construct 100(1 − α)% confidence intervals for the γ-quantile using procedures based on bootstrap, normal approximation and hypothesis testing. We show that the method based on hypothesis testing produces a confidence interval which is more satisfactory than those found by using bootstrap or normal approximation.