International Conference on Actuarial Sciences [ICActS] 2025
Permanent URI for this collectionhttps://rda.sliit.lk/handle/123456789/4496
Browse
Publication Open Access Quantifying Future Flood Risk in Sri Lanka: A Smart Data Approach for Insurance Pricing and Strategy(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Premaratne, CSri Lanka is increasingly vulnerable to flooding due to climate change, unplanned urban expansion, and insufficient infrastructure resilience. Despite this, the current insurance regulatory framework under the Risk-Based Capital (RBC) regime does not explicitly incorporate a catastrophic (CAT) risk charge for natural disasters such as floods. This paper proposes a novel framework for quantifying future flood risk in Sri Lanka using a smart data approach that integrates hydraulic modeling (HEC-RAS), geographic information systems (GIS), and machine learning (ML). The framework enables the generation of flood hazard maps, estimation of event probabilities, and calculation of expected losses at property level. A simulation-based approach is then used to determine the capital required to cover extreme loss events, which can serve as the basis for a CAT risk charge. Although full implementation is pending, this paperpresents an illustrative model using synthetic data to demonstrate the methodology and its potential implications. By embedding flood risk into pricing and strategic decisions, this approach aims to improve insurance sector resilience and inform regulatory advancement. The results highlight the feasibility and urgency of adopting data-driven tools to better manage climate-induced risks in Sri Lanka.Publication Open Access Study on Factors Affecting Purchase Intention of Fashion Clothes Advertised on Social Media Platforms(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Bandara, S. M.G. S.; Nethmini, K. A. D.; Nanayakkara,P. G. S. A.; Ranasinghe, H. M. D. K.; Niroshani, S. A.With the evolution of the digital world, social media has become a key player in the online fashion industry. This study on factors influencing purchase intention for fashion clothing through social media focused on how factors such as price, brand reputation, product quality, design variety, brand image, customer reviews, delivery time, return policy, and delivery quality impact purchasing decisions. A questionnaire featuring a five-point Likert scale was used to gather data. Using simple random sampling, 203 responses were collected within Sri Lanka. The importance of each factor was analysedby calculating the means from the Likert scale responses in MATLAB (R2018a). According to the Likert scale interpretation for a five-point scale, with (4.21–5.00) considered very important: price (4.66), product quality (4.79), delivery quality (4.59), variety of designs (4.23), customer reviews (4.43), delivery time (4.24), and return policy (4.28). Other factors, such as brand reputation (4.01) and brand image (3.63), fell within the important range (3.41–4.20). Overall satisfaction with online shopping (3.43) alsofalls within this range. Therefore, this study concludes that apparel businesses should focus their marketing strategies on these key factors via social media to improve customer engagement and increase sales.Publication Open Access Optimal Assignment of Machine Operators to Enhance Productivity and Skill Efficiency in Assembly Lines in Apparel Industry Using Two-Phase Integer Linear Programming Model: A Case Study(2025-10-10) Salay, M. I.; Senarathne, B. D. S. N.; Dissanayake, D. M. P. S.; Rajarathne, P. M. K.; Jayarathna, D. V. M.; Kularathna, A. P.; Kahadawa, J. D.; Daundasekara, D. M. W. B.Assembly line balancing (ALB) in the apparel industry is crucial for optimizing production efficiency. It requires an optimum assignment technique for machine operators (MOs) based on their skills and availability to minimize production delays and enhance productivity. This study aims to optimize the line layout and production throughput by applying a bottleneck-oriented resource allocation framework that combines the Rank Positional Weight Method (RPWM) with a two-phase IntegerLinear Programming model (ILPM). Once the line layout is determined by the RPWM, the next stage is to assign MOs to operations by solving a two-phase ILPM. In the first phase, an ILPM is applied to maximize the total production rate by assigning MOs to operations, based on their efficiency, identifying bottleneck operations which contribute to the lowest production rate. In second phase, the total skill level of the assembly line is minimized. The predetermined bottleneck production rate is used as an indicator, ensuring that the production rate which is maximized in the first phase is kept fixed.The reassignment of the remaining MOs is based on their skill levels, while the bottleneck operations and operators are kept aside in the second phase. The bottleneck operation, identified in the first phase, ensures that the most efficient MOs are assigned where needed, while other operations are conducted by MOs based on a compromise solution between their skill levels and availability. This approach emphasizes the importance of line balancing and operator assignment in the apparel industry and determining the ideal number of MOs needed to perform the set of operations. Additionally, thisproposed method can be adopted to any production line with necessary modifications.Publication Open Access Event Detection and Latency Analysis in High Frequency Trading Dashboards(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) de Silva, U; Perera, S; Liyanage, U.P; Erandi, HHigh frequency trading relies on millisecond-level decisions, where profitability is strongly influenced by both market responsiveness and system latency. Traditional dashboards offer real-time visualizations but fall short in detecting abrupt regime shifts or quantifying latency. This study presents an AI-aided Market Pulse and Latency Panel that integrates candlestick pattern recognition, change point detection and latency measurement into a unified dashboard. The system detects technical patterns, identifies structural market shifts, and quantifies infrastructural bottlenecks. Experimental results demonstrate that the panel enhances situational awareness by combining event detection with latency analytics, providing traders with actionable insights for strategy adjustment and infrastructural optimization.Publication Open Access Climate-Based Agri-Insurance Method for Paddy Production in Sri Lanka(Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Vijayakumar, JClimate change has emerged as a major threat to agriculture globally, and Sri Lanka is no exception. The districts of Ampara, Anuradhapura, and Polonnaruwa are the main producers of paddy. In recent decades, these regions have experienced greater climate variability, leading to unstable harvests and posing financial risks for paddy farmers. This study examines the potential of Weather Index Insurance (WII) as an effective tool to mitigate income losses caused by extreme weather events. Historical data on paddy yields were combined with daily weather records. The analysis focused on relationships between paddy yields and weather variables: total rainfall, average temperature, maximum and minimum temperatures, and extreme monthly temperatures in the regions for both the Maha and Yala cultivation seasons. Regression models identified significant correlations, and insurance indices were designed for each district and season, with pure premiums calculated based on these relationships. The results indicate that total rainfall is the most significant factor influencing yield variability across all three districts in different seasons. The proposed insurance models were able to reduce income variability by 15–19%. These findings indicate that rainfall is the most reliable basis for climate-resilient paddy insurance in these regions. This data-driven framework for index-based agricultural insurance provides insights to enhance farmer resilience, reduce economic vulnerability for farmers, and support the long-term sustainability of production in those regions.Publication Open Access A Machine Learning Approach to Actuarial Life Table Estimation in Lung Cancer Patients(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Tharushika, D. D. H.; Napagoda, N. A. D. N.Cancer-related mortalities worldwide are most caused by lung cancer, and one of the major causes of passing worldwide is still cancer. A dangerous disease is lung cancer, which requires accurate survival modelling to assist in actuarial evaluations, public health planning, and clinical decisions. Life expectancy and mortality risk across age groups are calculated using essential tools such as actuarial life tables, but complex real-world data is frequently struggled with by traditional methods. Actuarial life tables for patients with lung cancer are created using a data set of more than 500,000 patient records with 15 key variables from 2014 to 2024 across European countries, employing Extreme Gradient Boost Accelerated Failure Time (XGBoost AFT) based survival analysis. The main objective is to develop agespecific mortality rates and life expectancy for patients with lung cancer. In contrast to earlier research that was reliant on traditional models, the nonlinear learning capabilities of XGBoost AFT models areutilized in this study to allow for more accurate estimation of mortality trends. A data-driven, machine learning approach to actuarial life table development is contributed by this study, with information about lung cancer survival patterns being provided. The understanding of survival trends, treatment planning, efficient use of healthcare resources, and assessment of the results of initiatives is aided by physicians, researchers, and policymakers. Public health initiatives focused on early identification and prevention are also guided, as well as future healthcare requirements being forecast.Publication Open Access Latent Structures in Zero-Inflated Risk Domains: An Elastic–Tweedie Synergy for Claim Forecasting(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Kumarasinghe, P. B. W. S. R.; Napagoda, N. A. D. N.The frequency of insurance claims presents a unique modeling challenge due to high-dimensional inputs, strong feature correlations, and the dominance of zero-inflated outcomes. Conventional statistical models often fall short under these conditions, failing to capture the underlying structure of complex data sets. This study proposes an advanced predictive framework integrating Elastic Net regularization and a Tweedie-distribution-based XGBoost algorithm to address these issues in the context of motor insurance. Those methodologies were applied to the French Motor Claims data set,which contains over 678,000 policies, to distill influential variables while suppressing redundancy and noise. Lasso Regression, Elastic Net and the Boruta algorithm were employed to select relevant features. Elastic Net, in particular proved effective in identifying critical predictors including Exposure, Vehicle Age, Driver Age, BonusMalus, Area, and Fuel Type by balancing sparsity and multicollinearity. Thesefeatures were used to train both standard and Tweedie-distribution-based XGBoost models. Performance was evaluated using RMSE, MAE, and R², where the Tweedie XGBoost model guided by Elastic Net-selected features achieved the highest accuracy and explanatory power. The proposed architecture not only offers superior generalization and interpretability but also exhibits robustness in modeling skewed, zero-dominated distributions inherent to claim data. Beyond predictive enhancement, this framework has practical implications for actuarial science, particularly in dynamicpricing strategies, refined segmentation, and adaptive underwriting. This approach marks a shift toward more nuanced and scalable machine learning paradigms in insurance analytics by integrating statistically grounded feature selection with distribution-aware boosting.Publication Open Access Modelling the Indicative Rate of the USD/LKR SPOT Exchange Rate in Sri Lanka(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Rajapaksha, R. G. S. N.; Kumarasiri, P. V. A. L.; Sathsarani, T. V. I. A.; Rambukkana, P. P.; Botheju, W. S. R.; Guruge, M. L.; Peiris, T. S. G.This study develops and validates a time series model to forecast Sri Lanka’s daily indicative USD/LKR spot exchange rate using ARIMA and ARCH methods using data from 1st of January 2021 to 4th of June 2025, sourced from Central Bank of Sri Lanka. The original series was first differenced to achieve stationarity since it is not stationary. According to the sample ACF and PACF of stationary series, three candidate models were augmented with an ARCH(2) variance specification based on residual diagnostics. After comparing AIC, SIC, Hannan Quinn metrics and log likelihood, the ARIMA(1,1,1)+ARCH(2) was identified as the best possible model. The diagnostic tests confirmed that residuals are identically and independently distributed without remaining heteroskedasticity. Insample forecasting yielded a MAPE of 0.32% and a Theil U statistic of 0.0036, while out-of-sample validation (June 5 to July 4, 2025) produced a MAPE of 0.087% and a bias proportion near zero, highlighting the model the model’s predictive accuracy. By focusing only on the internal pattern of the exchange rate, this research creates a strong short term forecasting tool for Sri Lanka's volatile currencyenvironment laying ground work for adding outside factors in future improvements.Publication Open Access Predictive Model for the SPDR S&P 500 ETF (SPY) using Volatility Analysis Approach(Department of Mathematics and Statistics, Faculty of Humanities and Sciences,SLIIT, 2025-10-10) Musharraff, N. I.; Fernando, W. S. C.; Godage, T. R.; Jayasooriya, J. M. T. S.; Siriwardhana, H. A. A. T. P.; Samasundara, T. A.; Guruge, M. L.; Peiris, T. S. G.The S&P 500 (Standard & poor’s 500) is one of the most widely followed equity indices in the world. The SPDR S&P 500 ETF Trust (SPY) is used to track the performance of the S&P 500 index as closely as possible and can also be traded in the stock exchanges. Not many studies have been carried out to forecast daily closing prices of SPY for recent years. This study presents a time series analysis and forecasting of the daily closing prices of the SPY index. The dataset extends from 2000 to 2025, capturing key financial events, market movements and long-term growth trends. Due to high volatility, we were forced to consider variance equation in additional to the mean equation and the best fitted model identifies is ARIMA (1,1,1) + GARCH (1,1).ARIMAPublication Open Access Designing an Economic Scenario Generator for Financial Risk Management of Low-Income Households in Sri Lanka: A Review(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Peiris, K. G. H. S.; Premarathna, L. P. N. DLow-income households in Sri Lanka face increasing financial vulnerabilities driven by unstable income, high dependence on essential goods, and exposure to inflation and external shocks. Economic Scenario Generators (ESGs), widely used in institutional risk management, offer a structured way to model uncertainty but have rarely been adapted for household-level applications. This review synthesizes literature on ESG methodologies, household financial risk in developing economies, and Sri Lanka’s socio-economic realities. It highlights the need for a household-oriented ESG framework that integrates macroeconomic shocks with micro-level financial behavior to support budgeting, debt avoidance, and policy interventions.Publication Open Access Forecasting Monthly Electricity Consumption for Energy Planning and Policy Development in Sri Lanka(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Pathirana, D; Arachchige, C. N. P. GFor effective energy planning, grid stability, and policy development especially in emerging nations like Sri Lanka accurate electricity consumption projections is essential. The goal of this project is to create a reliable model that can forecast the Ceylon Electricity Board's (CEB) monthly electricity consumption using a large dataset that includes macroeconomic variables, market indicators, peak demand, energy generation sources, and weather data. Autoregressive Distributed Lag (ARDL), Random Forest, and eXtreme Gradient Boosting (XGBoost) are the three models whose redicting performance is compared in this study. The most pertinent predictors were chosen using Recursive Feature Elimination with Cross-Validation (RFECV). Although XGBoost performed well throughout training, overfitting was a problem. ARDL was interpretable, however it was unable to detect longterm cointegration and could not represent non-linear connections. With the best accuracy and dependability on the test dataset without overfitting, Random Forest turned out to be the best model whereas Monthly Sales by Tariff in LKR, Fuel Cost by Power Stations in LKR, Electricity Generation from Thermal Coal in CEB (Gwh), Electricity Generation from Mannar Wind in CEB (Gwh), Day Peak Demand (MW), Night Peak Demand (MW), Average Monthly Rainfall (mm), and Gross Domestic Product (GDP) in LKR were the eight most important factors that were found to be involved in forecasting electricity consumption. On the test dataset, Random Forest, the best model chosen, had an accuracy of 77.34%, a Mean Absolute Percentage Error (MAPE) of 22.67%, a Root Mean Square Error (RMSE) of 12.62, and a Mean Absolute Error (MAE) of 11.11. However, the models might not be able to reflect long-term structural changes like the switch to electric vehicles or widespread adoption of renewable energy sources, and the study did not account for new elements like government policy reforms or energy efficiency initiatives. Nevertheless, the results show that machine learning, in particular Random Forest, can improve Sri Lankan electricity consumption predictions to aid in sustainable energy planning and policy choices.Publication Open Access Development of Time Series Model to Predict the Weekly Percentage of Python Programming Language usage(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Gunawardane, D. M. N. M.; Herath, H. M. P. T.; Pitiyekumbura, W. S.; Samodhika, P. L. D.; Athauda, A. M. B. T.; Amarasinghe,E. J. C. U.; Peiris, T. S. G.Python's super popular and getting bigger fast. Figuring out how it will be used is super important for planning what to teach, training tech workers, and making good rules, especially in places like Sri Lanka that are just now getting into digital stuff. Therefore, this study aims to predict the weekly global usage of Python. We looked at data from April 21, 2019, to April 21, 2024. We got 262 weeks. This data is entered into Kaggle from Google search interest scores (Nextmillionaire, 2023). This dataset shows the highest interest score for Python in the general world. After trying out a bunch of models, theARIMA (1,1,1) model with seasonal stuff seemed like the best fit. We taught the model with data from April 21, 2019, to January 28, 2024 (250 weeks) and checked it with data from February 4, 2024, to April 21, 2024 (12 weeks). We tested the model to make sure it was doing things right, and the leftovers looked random, which is a good thing. The MAPE (Mean Absolute Percentage Error) for the validation data is 6.04%. This shows the ARIMA model is pretty good at guessing Python usage over time. Because theguesses are pretty accurate and consistent, it looks like Python usage of global is going up steadily. This means Python is a big deal with both Data Science & Analytics, Machine Learning & AI, Cloud Computing & DevOps, Automation & Scripting. This info should help schools, training places, and the government make smart choices about teaching digital skills.Publication Open Access Classifying Credit Card Payment Risk among Senior Citizens using Machine Learning on Limited Data(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Jayathilaka, G. A. M. K.; Ekanayake, E. M. P.; Appuhami, P. A. D. A. N.Sri Lanka is experiencing a significant demographic transition in view of lower fertility, enhanced life expectancy, and international migration, all of which have accounted for a higher proportion of senior citizens. The credit risk associated with a rising percentage of elderly population demands an investigation into the payment habits of senior citizens using credit facilities, for their own benefits and the sustainability of financial institutions. Addressing this issue, machine learning techniques are employed in this study in order to develop a viable model for classifying the credit card payment riskposed by the senior citizens based on their demographic and financial metrics at a leading private bank in Sri Lanka. Predictive dual-category classifications comprising of on-time payers and risky payers that include both late payers and dormant users are achieved using the machine learning algorithms of Logistic Regression, Random Forest, Support Vector Machine, Naïve Bayes, Extreme Gradient Boosting, and Multi-Layer Perceptron Neural Networks. Hyperparameter tuning and model performance ptimization were accomplished using Grid Search Cross-Validation, with models beingassessed by Accuracy, Precision, Recall, F1 Score, and AUC-ROC. Of those modelling techniques, the Random Forest excelled with 85% Accuracy, 85% Precision, 85.45% Recall, 84.99% F1 Score, and a high AUC-ROC score of 88 in classifying the risk posed by senior citizens, identifying payment settlement, average monthly credit card spending, average monthly debit, and number of credit card transactions as the key contributing variables. This classification method could be recommended for financial institutions with limited databases in setting credit limits tailored to senior citizens’ payment behaviour, reducing risk and promoting sustainable financial management for seniors.Publication Open Access A Poisson Mixture Model of Claim Counts to Improve Insurance Claim Predictions Using Incomplete Data/ Asymmetric Data: A Case Study with Telematics Insurance(2025-10-10) Peiris, K. G. H. S.; Sampath, J. K. H.; Premarathna, L. P. N. DIn the evolving landscape of insurance analytics, integrating traditional and telematics data is pivotal for enhancing the accuracy of claim predictions. This study introduces a two-fold approach utilizing a Poisson mixture model to merge these distinct data streams effectively. Initially, we apply the Poisson mixture model to traditional insurance features common to both datasets, employing Hamiltonian Monte Carlo (HMC) and Metropolis-Hastings algorithms separately for model fitting. Subsequently,the predicted claim counts derived from the Poisson mixture model are used as an offset to fit a Poisson generalized linear model (GLM) exclusively with telematics-based features. Our focus is on assessing the suitability of HMC and Metropolis-Hastings for addressing data integration challenges within Poisson mixture frameworks. Comparative analysis reveals that while HMC demands more computational time to achieve convergence, it exhibits superior performance in parameter estimation in scenarios with increased model complexity. This study underscores the potential of advanced Monte Carlo methods in refining predictive models by leveraging the synergy between traditional and telematics data sources.Publication Open Access A Novel Hypermatrix Product and its Application to Multilinear Mappings(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Senevirathne, S. S. M. A. C.; Athapattu, A. M. C. U. M.; Chathuranga, K. M. N. M.Matrix theory provides a well-established algebraic framework for working with linear maps, in which matrix multiplication replaces the composition of linear transformations. However, there is no canonical multiplication rule for hypermatrices that leads to multilinear maps, partly because multilinear maps are not closed under composition. To address this gap, this research introduces a novel (restricted) hypermatrix multiplication based on the Frobenius inner product. We start byshowing that every multilinear map 𝑓: 𝑉1 × 𝑉2 × … × 𝑉𝑛 → 𝑉0 gives a hypermatrix representation 𝒜 and defining a contraction operation, which computes 𝑓(𝑣1, 𝑣2, … , 𝑣𝑛 ) through Frobenius inner products between 𝒜 and matrices derived from input vectors. This operation allows for the efficient computation of the hypermatrix of an arbitrary multilinear map. This work provides constructive proofs and detailed numerical examples.Publication Open Access Predictive Model for Monthly Made Tea Production in Sri Lanka(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Subasinghe, C; Wattegedara, N; Silva, T; Balasooriya, S; Dassanayake, K; Guruge, M.LThis study forecasts monthly tea production in Sri Lanka by developing a suitable time series model to identify future trends in the national tea industry. The analysis is based on monthly made tea production data from January 2000 to June 2025, obtained from the Central Bank of Sri Lanka and the Sri Lanka Tea Board. After confirming the non-stationarity of the original series through the Augmented Dickey-Fuller test, both first-order and seasonal differencing were applied to achieve stationarity. The Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) plotswere used to identify potential model structures.Publication Open Access On the Fundamentals of Angle Trisectors of a Triangle(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Amarasinghe, IOver the years of the history of elementary and advanced geometry, trisecting a given angle into three equal parts, was prominent and given more attention. Nevertheless, it is evident that there is a significant research gap of the standard angle trisectors, the lengths of the angle trisectors and the relationships amongst other standard line segments in a triangle. In this paper, we address this gap by developing a purely geometric framework, supplemented with advanced algebraic methods, to obtainclosed-form expressions for internal angle trisectors in a Euclidean triangle. Using the circumcircle, similarity arguments, and Ptolemy’s Theorem, we derive polynomial relations and solve the associated cubic equations explicitly through Cardano’s method. The explicit determination of angle trisector lengths has not been previously available in closed form. Most approaches are trigonometric, but the trisector and related lengths were implicit or incomplete. Moreover, we present few very useful, novel, interesting lemmas, fundamental theorems, and corollaries related to two-dimensional angle trisectors in Euclidean triangles without using any trigonometric, vector algebra or complex number methods.Publication Open Access Sustainable Impact: Geo Pool Insure against Geo- Political Risk(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Krishnamoorthy, PThe Middle East is currently navigating a period of profound geopolitical flux and instability, with 2024 emerging as a year of significant regional transformation. Ongoing conflicts in Syria, Gaza, and Lebanon, exacerbated by the dramatic collapse of the Assad regime in late 2024, continue to reshape the political and humanitarian landscape. Recent escalations, such as Israel's "Operation Rising Lion" targeting Iranian nuclear facilities in June 2025, underscore the persistent risk of broader regional conflict. This volatility poses a direct threat to the stability of the Gulf Cooperation Council (GCC) states, given their critical energy infrastructure, vital shipping lanes, and the presence of U.S. military bases, all of which risk being drawn into any wider conflagration. In response, GCC states have pragmatically pursued de-escalation and rapprochement, including with Iran, and are increasingly adopting a non-aligned stance in global geopolitics to safeguard their economic survival and regional security. Amidst this geopolitical backdrop, the GCC countries host a substantial, yet often underrecognized, population of individuals displaced from war-torn nations. These populations, primarily originating from Syria, Yemen, Iraq, Afghanistan, and Sudan, have largely entered GCC states under labor migration frameworks, such as the Kafala system, rather than through formal refugee status. Data indicates that Saudi Arabia alone hosted approximately 745,580 Syrians in 2017, while the UAE had 50,463, Kuwait 142,000, and Qatar 12,320. Yemeni emigrants in 2019 included 750,919 in Saudi Arabia,202,574 in the UAE, 68,962 in Kuwait, and 35,574 in Qatar. The UAE hosts around 300,000 Afghans, and Saudi Arabia 132,282. Saudi Arabia is also a top destination for Sudanese emigrants. These displaced individuals, while contributing significantly to GCC economies, face profound vulnerabilities, including limited social protection, precarious legal status, and inadequate access to essential services. A critical gap exists in comprehensive life insurance and social protection for these vulnerable groups. Despite their vital economic contributions, migrant workers, including those displaced, often lack robust social safety nets, particularly coverage beyond work-related injuries. The Kafala system exacerbates this precarity by tying legal residency to employers, severely limiting workers' mobility and their practical access to services and legal recourse. Existing life insurance schemes for expatriates are frequently limited, expensive, or not tailored to the transient and vulnerable nature of many displaced populations. For instance, natural death, a common cause of mortality among blue-collar workers, is often not covered by mandatory employer insurance.Publication Open Access Exploring the Determinants of Medical Insurance Expenses: A Quantile Regression Approach(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Rathnayake, K; Somasiri, D; Abeygunawardana, T; Nugegoda, K; Fernando, N; Guruge, M. L.; Peiris, T. S. G.Healthcare insurance costs are influenced by a combination of biological and socioeconomic factors. This study investigates how age, body mass index (BMI), gender, and discount eligibility affect medical insurance expenses in the United States, using data from 1,338 individuals. Due to the right-skewed distribution of expenses, quantile regression was applied at the 25th, 50th, and 75th percentiles, providing insights across low-, medium-, and high-cost groups. Results show that age and BMI consistently increase insurance expenses, with stronger effects among high-cost patients. Genderdifferences also emerged, with females incurring higher costs than males at certain expenditure levels. Discount eligibility significantly reduced expenses across all quantiles. In contrast, the number of children was not a significant predictor and was excluded from the final model. Compared to ordinary least squares regression, quantile regression provided a more accurate assessment of cost determinants in skewed data. These findings highlight the importance of adopting advanced modeling approachesin insurance pricing and suggest that targeted policies addressing individuals having high BMI and equitable discount programs could improve healthcare affordability and risk management.Publication Open Access Automated Log Parsing and Anomaly Detection Using BERT and GPT-2: A Large Language Model Approach for IT Systems(Department of Mathematics and Statistics, Faculty of Humanities and Sciences, SLIIT, 2025-10-10) Sathyanjana, W. W. N. C.; Gunawardhane, H. M. K. T.; Kumara Samantha, B. T. G. S.; Perera, S.Logs are important for diagnosing and understanding the security and operations of IT systems. In these spectacles, the sheer volume of data and their inherent complexity do not allow for an isolated approach. Issues of scalability and adaptability majorly divest most rule-based systems in log analysis. This paper proposes an automatic approach that employs state-of-the-art Large Language Models to detect anomalies, suggest parsing templates, and improve log quality. The suggested system will try to integrate and analyse an Anomaly Detection module for identifying outliers and threats to security, a Pattern Recognition Engine for identifying semantic relations, and a Log Parsing Module for deriving structured patterns. All three collectively serve to enhance efficiency, adaptability, and real-time detection of the log analysis process. Before any LLM-based processing, the results of these preprocessing steps are put through tokenization and normalization. The system was evaluated in a combination of 16 log sources with over 32,000 entries. The model attained an accuracy of 96% in lassification; this shows that it performs well in identifying complex log structures and detecting anomalies. Compared to standard approaches, the framework reduces manual processes and increases interpretability on a large scale across diverse environments of IT. The paper describes a structured approach in AI-powered log analysis, which automates essential procedures to offer improved system reliability, as well as real-time security monitoring. Further directions include real-time streaminganalysis, addressing ethical concerns in log data processing, and enhancing explain ability.