Rajalingam Sokkalingam

Clustering is an important method in data analysis, faces challenges due to the different nature of datasets, resulting in certain algorithms being less effective and taking a long time. Choosing the most effective clustering method involves evaluating its accuracy and computational speed for a dataset poses a significant challenge for today's researchers. To address these issues, current study compares different clustering methods, by using datasets, including iris, seed, and well log to evaluate their accuracy and execution speed. Results show that K-means performs better with large datasets. As sample size increases, the accuracy of the K-means algorithm tends to improve. The execution time of k-means is influenced by the number of features in the dataset, with datasets having a larger number of features typically requiring more time to process. Mean shift algorithm and spectral clustering algorithm are performed well in small data sets, but it takes a long time.

Accurate forecasting of crude oil prices is crucial for informed decision-making in global energy markets, financial planning, and policy formulation. This study examines the effectiveness of regression models in forecasting crude oil prices using historical data augmented with lagged features to capture temporal dynamics. Four models were evaluated, including gradient boosting, random forest, support vector regression (SVR), and a voting regressor ensemble that integrates predictions from both random forest and gradient boosting. The dataset was transformed using a seven-day lag structure, and model performance was evaluated using root mean squared error (RMSE), mean absolute error (MAE), and the coefficient of determination (R²). Results indicate that the voting regressor ensemble outperformed individual models, achieving the best performance metrics (RMSE = 3.5978, MAE = 1.9816, R² = 0.9363). Random forest and gradient boosting models also demonstrated robust performance (R² > 0.93). In contrast, SVR demonstrated relatively lower performance with an R² of 0.5614 and significantly higher error metrics. The results confirm that the ensemble voting regressor provides superior prediction accuracy for crude oil price forecasting. This finding has important implications for financial analysts, policymakers, and energy sector stakeholders aiming to improve decision-making based on reliable crude oil price forecasts.

High-dimensional sparse numerical data are normally encountered in machine learning, recommender systems, finance and medical imaging. The problem with this type of data is that it has high dimensions (many features) and highly sparse (most values are zero), which is prone to overfitting. The data visualization can be achieved through a neural network architecture called stacked autoencoders. These multilayer autoencoders are designed to reconstruct input data, but overfitting is a major problem. To overcome this problem novel L1 Regularization-dropout technique is introduced to reduce overfitting and boost stacked autoencoder performance. L1 regularization penalizes large weights, simplifying data representations whereas the dropout technique randomly turns off neurons during training and makes the model dependent only on the selected turn-on neurons. The model employs batch normalization to improve the performance of the autoencoder. The approach was implemented on a high-dimensional sparse numerical dataset in the field of cybersecurity to minimize the loss function, measured by Mean Square Error (MSE) and Mean Absolute Error (MAE). The findings were compared to the conventional stacked autoencoder. The study revealed that the suggested method effectively mitigated the issue of overfitting. Stacked autoencoders, when combined with L1 regularisation and the dropout approach, are very successful in handling high-dimensional sparse numerical data in a diverse range of applications.

Stochastic modeling approaches have attracted many researchers to the field. However, fire hotspot detection suffers from not using a Markov chain quasi-Monte Carlo (MCQMC) as a forecasting methodology. This paper proposes improvements to the computational time by combining the strengths of the Markov chain Monte Carlo (MCMC) and quasi-Monte Carlo (QMC) methods. The proposed method can lead to more precise and stable results, particularly in problems with high-dimensional integration or complex probability distributions. The proposed method is applied to a case study of fire hotspot detection in Sarawak, Malaysia. The outcome of this study reveals that the MCQMC method is more computationally efficient, taking only 0.0746 seconds compared to MCMC’s 0.0914 seconds and QMC’s 0.0994 seconds. It is shown that the best option derived by the proposed method is effective in predicting fire hotspots and providing quick solutions to protect the environment and communities from forest fires.

In this study the new statistical method kernelized rank order (KROD) numerical solution is analyzed to convert non spherical data to spherical form using diffe In this study the new statistical method kernelized rank order (KROD) numerical solution is analyzed to convert non spherical data to spherical form using diffe

The Burr III distribution is extended in this work as a substitute for the numerous Burr III distributions. A new distribution is developed by applying the log transformation technique to define the transformed log-Burr III distribution. Moments and quantile function are the structural features established in this study. The model parameters are derived using the maximum likelihood technique. The applicability of the new distribution was assessed using real-world data on the transformed total milk production in the first birth of 107 cows of the SINDI race. The results showed that the proposed distribution might be used as the optimal distribution for this dataset.

The quest for energy efficiency in buildings has placed a demand for designing and modeling energy-efficient buildings. In this study, the thermal energy performance of a tropically adapted passive building was investigated in the warm tropical climate of Malaysia. Two mock-up buildings were built to represent a “green”, made of clay brick double-glazed passive building and a conventional, made of concrete “red” building. The mean indoor temperature of the passive building was found to be always lower than that of the red building throughout the experiment during different weather constellations. Our research builds upon existing work in the field by combining multiple linear regression models and distribution models to provide a comprehensive analysis of the factors affecting the indoor temperature of a building. The results from the fitted multiple linear regression models indicate that walls and windows are critical components that considerably influence the indoor temperature of both passive buildings and red buildings, with the exception of passive buildings during the hot season, where the roof has a greater influence than the window. Furthermore, the goodness-of-fit test results of the mean indoor temperature revealed that the Fréchet and Logistic probability models fitted the experimental data in both cold and hot seasons. It is intended that the findings of this study would help tropical countries to devise comfortable, cost-effective passive buildings that are green and energy efficient to mitigate global warming.

In parametric statistical modeling, it is important to construct new extensions of existing probability distributions (PDs) that can make modeling data more flexible and help stakeholders make better decisions. In the present study, a new family of probability distributions (FPDs) called the odd beta prime generalized (OBP-G) FPDs is proposed to improve the traditional PDs. A new PD called the odd beta prime-logistic (OBP-logistic) distribution has been developed based on the developed OBP-G FPDs. Some desirable mathematical properties of the proposed OBP-logistic distribution, including the moments, moment-generating function, information-generating function, quantile function, stress–strength, order statistics, and entropies, are studied and derived. The proposed OBP-logistic distribution’s parameters are determined by adopting the maximum likelihood estimation (MLE) method. The applicability of the new PD was demonstrated by employing three data sets and these were compared by the known extended logistic distributions, such as the gamma generalized logistic distribution, new modified exponential logistic distribution, gamma-logistic distribution, exponential modified Weibull logistic distribution, exponentiated Weibull logistic distribution, and transmuted Weibull logistic distribution. The findings reveal that the studied distribution provides better results than the competing PDs. The empirical results showed that the new OBP-logistic distribution performs better than the other PDs based on several statistical metrics. We hoped that the newly constructed OBP-logistic distribution would be an alternative to other well-known extended logistic distributions for the statistical modeling of symmetric and skewed data sets.

In this work, a new bivariate inverted Nakagami distribution that can be used to model real-world datasets has been investigated. The newly developed bivariate distribution's cumulative distribution function and probability density function are defined. The bivariate distribution derives from the Farlie Gumbel Morgenstern, and the marginal density functions are also determined. Some fundamental estimation techniques, such as maximum-likelihood estimation and inference functions for margins, are used to derive the parameters of its estimates. Applications to real-world datasets pertaining to kidney infection diseases and the UEFA Champions' League group stage for the seasons 20042005 and 20052006 help to assess the efficacy of the proposed distribution.

Biohydrogen production from microalgae is a potential alternative energy source that is now intensively being researched. The complex natures of the biological processes involved have afflicted the accuracy of traditional modelling and optimization, besides being costly. Accordingly, machine learning algorithms have been employed to overcome setbacks, as these approaches have the capability to predict nonlinear interactions and handle multivariate data from microalgal biohydrogen studies. Thus, the review focuses on revealing the recent applications of machine learning techniques in microalgal biohydrogen production. The working principles of random forests, artificial neural networks, support vector machines, and regression algorithms are covered. The applications of these techniques are analyzed and compared for their effectiveness, advantages and disadvantages in the relationship studies, classification of results, and prediction of microalgal hydrogen production. These techniques have shown great performance despite limited data sets that are complex and nonlinear. However, the current techniques are still susceptible to overfitting, which could potentially reduce prediction performance. These could be potentially resolved or mitigated by comparing the methods, should the input data be limited.