Browsing by Author "Hewapathirana, Isuru"

Now showing 1 - 3 of 3

Modeling and Forecasting Mortality in Sri Lanka
(2014) Aberathna, Wasana; Alles, Lakshman; Wickremasinghe, W. N.; Hewapathirana, Isuru
The purpose of this study is to develop sex-specific mortality estimation models using historical mortality data for Sri Lanka, based on the statistical time series techniques attributed to Lee and Carter (1992). Historical mortality data was analyzed in the light of significant historical episodes. Several alternative univariate time series models were examined for modeling males and females, as well as a bivariate vector autoregressive (VAR) model. The VAR model when fitted to the first differenced series performed better than the univariate models and hence used for forecasting purposes. From the estimated VAR model, mortality forecasts were generated for the period up to 2030 and life tables were generated for the selected periods of 2006-2008.
Simulation analysis of an expressway toll plaza
(Department of Industrial Management, Faculty of Science, University of Kelaniya Sri Lanka, 2021) Grabau, Shehara; Hewapathirana, Isuru
Since the early civilizations, transportation has played a significant role, from fulfilling basic human needs to contributing towards major economic growths all over the world. With the advancement in technology, the demand for smooth and hassle-free transportation increased and it is particularly true for road transportation in Sri Lanka as well. As a result, the expressway road network was introduced to Sri Lanka in 2011. Although a toll is payable for the use of expressways, many vehicle users prefer to utilize the expressway due to the extensive amount of time saved. Time is of utmost importance for expressway users. Hence, long queues and waiting time at toll plazas where the toll payment is made should be minimized. This study is aimed at analyzing the performance at the Peliyagoda toll plaza of the Colombo-Katunayake expressway where the formation of long queues and long waiting time in queues can be observed during peak hours. Due to the high complexity of using the analytical approach in obtaining the performance measures, a simulation approach was used with Arena Simulation Software. Few setup improvements were identified, and each of the setups were simulated to obtain the performance measures. Based on the comparison of the results, recommendations and suggestions to improve the efficiency of the operations at the Peliyagoda toll plaza have been outlined.
A Systematic Investigation on the Effectiveness of the Tabbert Model for Credit Card Fraud Detection
(Department of Industrial Management, Faculty of Science, University of Kelaniya Sri Lanka, 2022) Hewapathirana, Isuru; Kekayan, Nanthakumar; Diyasena, Deshanjali
As a result of rapid digitisation, online transactions using credit cards have become popular. With this, fraudulent activities have also increased considerably. Although many supervised and unsupervised machine learning techniques were proposed in past research for identifying fraudulent transactions, they do not fully utilize the tabular and hierarchical structure present in transaction datasets. Recently, the TabBERT neural network model was proposed to calculate row-wise embeddings that capture both inter and intra dependencies between transactions in tabular time series data. In this research, we present a systematic experimental framework to assess the effectiveness of applying the embeddings calculated using the TabBERT model for credit card fraud detection. We employ the calculated row embeddings for fraud detection using three unsupervised machine learning algorithms and two supervised machine learning algorithms. We perform our experiments on a synthetic dataset that has been generated using the TabGPT model. Overall, TabBERT-based embeddings increase the performance of the supervised learning models with the extreme gradient boosting model achieving a precision of 99% and an F1 score of 98%, and the multilayer neural network model achieving a precision of 97% and an F1 score of 95%. For unsupervised learning, the use of TabBERT embeddings increases the recall rate of K-means clustering algorithm by 0.19%.