arXiv subjects

Thiyanga S. Talagala

Publications and source records attributed to Thiyanga S. Talagala.

13 recordsLinked to original sources

arXiv2025

quollr: An R Package for Visualizing 2-D Models from Nonlinear Dimension Reductions in High-Dimensional Space

Nonlinear dimension reduction methods provide a low-dimensional representation of high-dimensional data by applying a Nonlinear transformation. However, the complexity of the transformations and data structures can create wildly different representations depending on the method and hyper-parameter choices. It is difficult to determine whether any of these representations are accurate, which one is the best, or whether they have missed important structures. The R package quollr has been developed as a new visual tool to determine which method and which hyper-parameter choices provide the most accurate representation of high-dimensional data. The scurve data from the package is used to illustrate the algorithm. Single-cell RNA sequencing (scRNA-seq) data from mouse limb muscles are used to demonstrate the usability of the package.

stat.ME↗

arXiv2025

cardinalR: Generating Interesting High-Dimensional Data Structures

Simulated high-dimensional data is useful for testing, validating, and improving algorithms used in dimension reduction, supervised and unsupervised learning. High-dimensional data is characterized by multiple variables that are dependent or associated in some way, such as linear, nonlinear, clustering or anomalies. Here we provide new methods for generating a variety of high-dimensional structures using mathematical functions and statistical distributions organized into the R package cardinalR. Several example data sets are also provided. These will be useful for researchers to better understand how different analytical methods work and can be improved, with a special focus on nonlinear dimension reduction methods. This package enriches the existing toolset of benchmark datasets for evaluating algorithms.

stat.ME↗

arXiv2025

Choosing Better NLDR Layouts by Evaluating the Model in the High-dimensional Data Space

Nonlinear dimension reduction (NLDR) techniques such as tSNE, and UMAP provide a low-dimensional representation of high-dimensional data ($p\text{-}D$) by applying a nonlinear transformation. NLDR often exaggerates random patterns. But NLDR views have an important role in data analysis because, if done well, they provide a concise visual (and conceptual) summary of $p\text{-}D$ distributions. The NLDR methods and hyper-parameter choices can create wildly different representations, making it difficult to decide which is best, or whether any or all are accurate or misleading. To help assess the NLDR and decide on which, if any, is the most reasonable representation of the structure(s) present in the $p\text{-}D$ data, we have developed an algorithm to show the $2\text{-}D$ NLDR model in the $p\text{-}D$ space, viewed with a tour, a movie of linear projections. From this, one can see if the model fits everywhere, or better in some subspaces, or completely mismatches the data. Also, we can see how different methods may have similar summaries or quirks.

stat.ME↗

arXiv2025

Relationship Between Leisure Activities, Stress Management Methods, Study Methods, and Methods of Learning New Things Among First-Year Statistics Students

The interplay between leisure activities, stress management methods, studying methods, and methods of learning new things is crucial and affects performance in all aspects of life. On the other hand, data science and statistics are rapidly growing fields with high demands across universities. Thus, this study aimed to identify the similarities and dissimilarities between the four dimensions: leisure activities, stress management methods, studying methods and methods of learning new things. The participants of this study were first-year undergraduates studying statistics at one of the universities in Sri Lanka. There were 117 students in the sample (female-65, male-52). A self-reported questionnaire was used to collect data. First, individual responses for each question under each dimension were visualized using tile maps separately for males and females to identify similarities and dissimilarities in responses. Next, individuals were clustered based on the responses for each dimension separately. Finally, all resulting clusters were re-clustered to identify the relationships between the dimensions. In all cluster analyses, we used Jaccard distance with hierarchical clustering using the complete linkage method. The results were visualized using tile maps. Across all four dimensions we considered, the top activities were either listening to music or lectures and watching videos or TV shows, suggesting that individuals are introverts and passive learners. There was no strong relationship between these dimensions. By identifying these clusters and relationships, educators can tailor instructional approaches to enhance engagement and effectiveness in diverse learning environments.

stat.OT↗

arXiv2024

tsdataleaks: An R Package to Detect Potential Data Leaks in Forecasting Competitions

Forecasting competitions are of increasing importance as a means to learn best practices and gain knowledge. Data leakage is one of the most common issues that can often be found in competitions. Data leaks can happen when the training data contains information about the test data. There are a variety of different ways that data leaks can occur with time series data. For example: i) randomly chosen blocks of time series are concatenated to form a new time series; ii) scale-shifts; iii) repeating patterns in time series; iv) white noise is added to the original time series to form a new time series, etc. This work introduces a novel tool to detect these data leaks. The tsdataleaks package provides a simple and computationally efficient algorithm to exploit data leaks in time series data. This paper demonstrates the package design and its power to detect data leakages with an application to forecasting competition data.

stat.AP↗

arXiv2024

ceylon: An R package for plotting the maps of Sri Lanka

The rapid evolution in the fields of computer science, data science, and artificial intelligence has significantly transformed the utilisation of data for decision-making. Data visualisation plays a critical role in any work that involves data. Visualising data on maps is frequently encountered in many fields. Visualising data on maps not only transforms raw data into visually comprehensible representations but also converts complex spatial information into simple, understandable form. Locating the data files necessary for map creation can be a challenging task. Establishing a centralised repository can alleviate the challenging task of finding shape files, allowing users to efficiently discover geographic data. The ceylon R package is designed to make simple feature data related to Sri Lanka's administrative boundaries and rivers and streams accessible for a diverse range of R users. With straightforward functionalities, this package allows users to quickly plot and explore administrative boundaries and rivers and streams in Sri Lanka.

stat.OT↗

arXiv2023

scatteR: Generating instance space based on scagnostics

Traditional synthetic data generation methods rely on model-based approaches that tune the parameters of a model rather than focusing on the structure of the data itself. In contrast, Scagnostics is an exploratory graphical method that captures the structure of bivariate data using graph-theoretic measures. This paper presents a novel data generation method, scatteR, that uses Scagnostics measurements to control the characteristics of the generated dataset. By using an iterative Generalized Simulated Annealing optimizer, scatteR finds the optimal arrangement of data points that minimizes the distance between current and target Scagnostics measurements. The results demonstrate that scatteR can generate 50 data points in under 30 seconds with an average Root Mean Squared Error of 0.05, making it a useful pedagogical tool for teaching statistical methods. Overall, scatteR provides an entry point for generating datasets based on the characteristics of instance space, rather than relying on model-based simulations.

stat.CO↗

arXiv2022

Interactive Dashboard to Monitor the COVID-19 Outbreak and Vaccine Administration

Dashboards are the most common visualization method for displaying COVID-19 data and informing the public. We examined 15 different dashboards to see how various visualization techniques were used. This paper describes the creation and implementation of a dashboard for COVID-19 epidemic and vaccination administration data in Sri Lanka.

stat.AP↗

arXiv2022

Forecasting: theory and practice

Forecasting has always been at the forefront of decision making and planning. The uncertainty that surrounds the future is both exciting and challenging, with individuals and organisations seeking to minimise risks and maximise utilities. The large number of forecasting applications calls for a diverse set of forecasting methods to tackle real-life challenges. This article provides a non-systematic review of the theory and the practice of forecasting. We provide an overview of a wide range of theoretical, state-of-the-art models, methods, principles, and approaches to prepare, produce, organise, and evaluate forecasts. We then demonstrate how such theoretical concepts are applied in a variety of real-life contexts. We do not claim that this review is an exhaustive list of methods and applications. However, we wish that our encyclopedic presentation will offer a point of reference for the rich work that has been undertaken over the last decades, with some key insights for the future of forecasting theory and practice. Given its encyclopedic nature, the intended mode of reading is non-linear. We offer cross-references to allow the readers to navigate through the various topics. We complement the theoretical concepts and applications covered by large lists of free or open-source software implementations and publicly-available databases.

stat.AP↗

arXiv2021

COVID-19 and Online Learning Tools

Distance education has a long history. However, COVID-19 has created a new era of distance education. Due to the increasing demand, various distance learning solutions have been introduced for different distance education purposes. In this study, we investigated the impact of COVID-19 on global attention towards different distance learning-teaching tools. We used Google Trend search queries as a proxy to quantify the popularity and public interest towards different distance education solutions. Both visual and analytical approaches were used to analyze global-level web search queries during the COVID-19 pandemic. This can provide a fast first step guide to identifying the most popular online learning tools available for different educational purposes. The results allow the teachers to narrow down the search space and deepen their exploration of prominent distance education solutions to support their online teaching. The R code and data to reproduce the results of this work are available in the online supplementary materials.

cs.CY↗

arXiv2021

Hierarchical Forecasting of Dengue Incidence in Sri Lanka

The recurrent thread of dengue incidence in Sri Lanka is still abundant and it creates a huge burden to the country. Hence, the National Dengue Control Unit of Sri Lanka propose a national action plan to prevent and control the dengue incidence. To implement the necessary actions for short and long terms the proposed plan operates under three levels:country-level, province-level, and district levels. In order to optimize resource allocation, the health officers require the forecasts for country, province and district levels, which preserves the aggregate consistency associated with the district, province, and country levels as well as time correlations. Hence, the objective of this study is to forecast the dengue incidence in Sri Lanka using a hierarchical time series forecasting approach based on spatial and temporal hierarchical structures. Hierarchical forecasting involves two steps such as generating base forecasts and reconciliation of these base forecasts. In this study, Exponential smoothing(ETS) and Autoregressive Integrated Moving Average (ARIMA), NAIVE, Seasonal NAIVEapproaches and average method are used to generate base forecasts. Accuracy of forecasts is evaluated using Mean Absolute Scale Error (MASE). We compare the accuracy of hierarchical time series forecasts with other benchmark approaches. The forecast accuracy reveals that the best forecasting approaches for country, provinces and districts are not limited to a single approach. Hence, we investigate reasons for the variations of performance in different forecasting approaches based on a time series feature-based visualization approach.

stat.AP↗

arXiv2021

Computer-aided Interpretable Features for Leaf Image Classification

Plant species identification is time consuming, costly, and requires lots of efforts, and expertise knowledge. In recent, many researchers use deep learning methods to classify plants directly using plant images. While deep learning models have achieved a great success, the lack of interpretability limit their widespread application. To overcome this, we explore the use of interpretable, measurable and computer-aided features extracted from plant leaf images. Image processing is one of the most challenging, and crucial steps in feature-extraction. The purpose of image processing is to improve the leaf image by removing undesired distortion. The main image processing steps of our algorithm involves: i) Convert original image to RGB (Red-Green-Blue) image, ii) Gray scaling, iii) Gaussian smoothing, iv) Binary thresholding, v) Remove stalk, vi) Closing holes, and vii) Resize image. The next step after image processing is to extract features from plant leaf images. We introduced 52 computationally efficient features to classify plant species. These features are mainly classified into four groups as: i) shape-based features, ii) color-based features, iii) texture-based features, and iv) scagnostic features. Length, width, area, texture correlation, monotonicity and scagnostics are to name few of them. We explore the ability of features to discriminate the classes of interest under supervised learning and unsupervised learning settings. For that, supervised dimensionality reduction technique, Linear Discriminant Analysis (LDA), and unsupervised dimensionality reduction technique, Principal Component Analysis (PCA) are used to convert and visualize the images from digital-image space to feature space. The results show that the features are sufficient to discriminate the classes of interest under both supervised and unsupervised learning settings.

cs.CV↗

arXiv2021

FFORMPP: Feature-based forecast model performance prediction

This paper introduces a novel meta-learning algorithm for time series forecast model performance prediction. We model the forecast error as a function of time series features calculated from the historical time series with an efficient Bayesian multivariate surface regression approach. The minimum predicted forecast error is then used to identify an individual model or a combination of models to produce the final forecasts. It is well-known that the performance of most meta-learning models depends on the representativeness of the reference dataset used for training. In such circumstances, we augment the reference dataset with a feature-based time series simulation approach, namely GRATIS, in generating a rich and representative time series collection. The proposed framework is tested using the M4 competition data and is compared against commonly used forecasting approaches. Our approach provides comparable performances to other model selection/combination approaches but at a lower computational cost and a higher degree of interpretability, which is important for supporting decisions. We also provide useful insights regarding which forecasting models are expected to work better for particular types of time series, the intrinsic mechanisms of the meta-learners and how the forecasting performances are affected by various factors.

stat.AP↗