Information Systems and Technology

Permanent URI for this collection

https://hdl.handle.net/10315/27588

Browse

Now showing 1 - 13 of 13

Open Access
A Hybrid Approach for Large-Scale Product Categorization Based on Weighted KNN and LSTM-BPV
(2019-12-04) Hu, Haohao; Huang, Xiangji
In modern e-commerce systems, large volumes of new items are being added to the product list everyday, which calls for automatic product categorization. In this thesis we propose a weighted K-Nearest Neighbour (KNN) based classification system for solving large-scale e-commerce product taxonomy classification problem. We use information retrieval (IR) model as similarity function in our weighted KNN algorithm. Among all IR models used in this study, we achieved highest classification performance through using information-based (IB) model as similarity function in the KNN algorithm. Moreover, our proposed method can improve the overall performance when combining prediction results with those from advanced neural network based method, namely Long Short-Term Memory with Balanced Pooling Views (LSTM-BPV). The hybrid system could achieve results comparable to the state of the art (SotA). We also get good results by fine-tuning pre-trained Bidirectional Encoder Representations from Transformers (BERT) model.
Open Access
Automatic Image Recognition of Rapid Malaria Emergency Diagnosis: A Deep Neural Network Approach
(2018-03-01) Liang, Zhaohui; Huang, Xiangji
Deep learning is the state-of-the-art artificial intelligence (AI) method for visual pattern detection and automated diagnosis. This paper describes the application of convolutional neural network (CNN), the deep learning model for visual recognition, to automatic detection of plasmodium parasitized red blood cells for malaria field screening and rapid diagnosis. The malaria thin blood smears are from Bangladesh and initially labeled by a specialist. 27,578 red blood cell images are segmented (raw set). The images are rotated clockwise three times to generate an augmented dataset with 110,312 red blood cell images. A 12-layer and an 18-layer CNN-based Malaria Net models are applied to classify both the raw data set and the augmented dataset. The performance is evaluated by ten-fold cross-validation and compared to a transfer learning model. In the ten-fold cross-validation test for Malaria Net, the average accuracy is 97.37% (18-layer) and 96.09% (12-layer) with the raw set, and is 97.93% and 96.75% with the augmented set, in comparison to 91.99% with the raw set and 94.26% with the augmented set in transfer learning. In addition, the two CNN models show superiority over transfer learning in all performance indicators such as sensitivity, specificity, precision, F1 score, and Matthews correlation coefficient. The Malaria Net can accurately detect malaria-infected red blood cells. A CNN model trained by domain-specific data shows superior performance over the transfer-learning method. Automatic image classification powered by deep learning offers not only an accurate method for the malaria field screening and rapid diagnosis but also a new solution for malaria control especially in resource-poor regions.
Open Access
Comparative Analysis of Language Models on Augmented Low-Resource Datasets for Application in Question & Answering Systems
(2024-11-07) Ranjbargol, Seyedehsamaneh; Erechtchoukova, Marina G.
This thesis aims to advance natural language processing (NLP) in question-answering (QA) systems for low-resource domains. The research presents a comparative analysis of several pre-trained language models, highlighting their performance enhancements when fine-tuned with augmented data to address several critical questions, such as the effectiveness of synthetic data and the efficiency of data augmentation techniques for improving QA systems in specialized contexts. The study focuses on developing a hybrid QA framework that can be integrated with a cloud-based information system. This approach refines the functionality and applicability of QA systems, boosting their performance in low-resource settings by using targeted fine-tuning and advanced transformer models. The successful application of this method demonstrates the significant potential for specialized, AI-driven QA systems to adapt and thrive in specific environments.
Open Access
Comparative Analysis of Transformer-Based Language Models for Text Analysis in the Domain of Sustainable Development
(2023-08-04) Safwat, Nabil; Erechtchoukova, Marina G.
With advancements of Artificial Intelligence, Natural Language Processing (NLP) has gained a lot of attention because of its potential to facilitate complex human-machine interactions, enhance language-based applications, and automate processing of unstructured texts. The study investigates the transfer learning approach on Transformer-based Language models, abstractive text summarization approach, and their application to the domain of Sustainable Development with the goal to determine SDGs representation in scientific publications using the text summarization technique. To achieve this, the traditional transfer learning framework was expanded so that: (1) the relevance of textual documents to specified text can be evaluated, (2) neural language models, namely BART and T5, were selected, and (3) 8 text similarity measures were investigated to identify the most informative ones. Both the BART and T5 models were fine-tuned on an acquired domain-specific corpus of scientific publications extracted from Scopus Elsevier database. The relevance of recently published works to an SDG was determined by calculating semantic similarity scores between each model generated summary to the SDG’s description. The proposed framework made it possible to identify goals that dominated the developed corpus and those that require further attention of the research community.
Open Access
Data-Driven Causal Decision Support for Business Process Management
(2024-07-18) Jandaghi Alaee, Ali; Senderovich, Arik
Control-flow and resource assignment decisions influence business processes. Recorded process data can be used to identify which decisions are informed by data to predict their outcome, and to guide interventions as part of a what-if analysis. The latter requires causal models that explain decisions. Yet, existing methods are limited: they focus on control-flow decisions only, ignore potential confounders, and use ad-hoc methods to resolve causal conflicts. We fill this gap, by introducing a causal decision modeling framework which uncovers confounding effects, and captures resource decisions. Moreover, we provide a process-aware causal discovery algorithm that takes process precedence into account. In addition, we employ domain knowledge to include unobserved factors. We address the problem of identification, conduct interventional outcome prediction and improve decision-making by acquiring unavailable data to maximize the utility of interventions. We demonstrate the feasibility of our approach through a set of experiments on synthetically generated and real-world datasets.
Open Access
Dynamic Elastic Provisioning For NFV-Enabled 5G Networks Using Machine Learning
(2023-03-28) Ali, Khalid; Jammal, Manar
5G networks are expected to support a variety of services and applications by having a more stringent latency, reliability, and bandwidth requirements compared to previous generations. To meet these requirements, Open Radio Access Networks (O-RAN) has been proposed. The O-RAN Alliance assumes O-RAN components to be Virtualized Network Functions (VNFs). Furthermore, O-RAN allows employing Machine Learning (ML) solutions to tackle challenges in resource management. However, intelligently managing resources for O-RAN can prove challenging. Network providers need to dynamically scale resources in response to incoming traffic. Elastically allocating resources provides higher flexibility, reduces OPerational EXpenditure (OPEX), and increases resource utilization. In this work, we propose and evaluate an elastic VNF orchestration framework for O-RAN. The proposed system consists of a traffic forecasting-based dynamic scaling scheme using ML, and a Reinforcement Learning (RL) based VNF placement policy. The models are evaluated based on their predictive capabilities subject to all Service-Level Agreements.
Open Access
Efficient Calculation of Optimal Configuration Processes
(2015-12-16) Fernandez, Yasser Gonzalez; Chen, Stephen; Liaskos, Sotirios
Customers are getting increasingly involved in the design of the products and services they choose by specifying their desired characteristics. As a result, configuration systems have become essential technologies to support the development of mass-customization business models. These technologies facilitate the configuration of complex products and services that otherwise could generate many incorrect configurations and overwhelm users with confusion. This thesis studies the problem of optimizing the user interaction in a configuration process – as in minimizing the number of questions asked to a user in order to obtain a fully-specified product or service configuration. The work carried out builds upon a previously existing framework to optimize the process of configuring a software system, and focuses on improving its efficiency and generalizing its application to a wider range of configuration domains. Two solution methods along with two alternative ways of specifying the configuration models are proposed and studied on different configuration scenarios. The experimental study evidences that the introduced solutions overcome the limitations of the existing framework, resulting in more suitable algorithms to work with models involving a large number of configuration variables.
Open Access
Enhancing General Language Models for Biomedical Test Retrieval via Diversified Prior Knowledge
(2023-12-08) Huang, Yizheng; Huang, Jimmy
The thesis introduces the Diversified Prior Knowledge Enhanced General Language Model (DPK-GLM) to improve the efficacy of general language models in biomedical Information Retrieval (IR). General language models often struggle with biomedical data due to its specialized terminology and the need for precise matching. DPK-GLM tackles these challenges by integrating domain-specific knowledge, thereby enhancing the model's ability to understand and process biomedical information. The framework comprises three core components. The first, Knowledge-based Query Expansion, leverages authoritative biomedical databases to enrich search queries with domain-specific entities. The second, Aspect-based Filter, identifies documents that are highly relevant to the query. The third, Diversity-based Score Reweighting, re-ranks these filtered documents by combining similarity and diversity scores, yielding more accurate results. Experimental tests on public biomedical IR datasets confirm that DPK-GLM significantly improves retrieval performance.
Open Access
Implementing Security Requirements through Automatic Generation of Secure Workflows
(2022-08-08) Jaouhar, Ibrahim; Liaskos, Sotirios
Modern software-intensive information systems are enormously large and complex. Prior to the design process of such systems, designers and architects need to know what kinds of stakeholder needs the system is supposed to support. This is particularly true for security requirements which must be captured and analyzed alongside all other requirements rather than treated as an afterthought. Hence, many researchers have proposed different modelling frameworks in different domain fields to address security and privacy patterns. However, most of these frameworks focus on comprehensive representation and analysis of requirements, without indicating how such requirements can be implemented within the context of a business process. Users are often at loss with regards to what security technologies they should adopt and incorporate in their workflows to reach secure business processes. In this thesis, we propose a framework for enriching goal-oriented requirements models with security controls necessitated by specified security requirements. A set of patterns are designed by security experts that associate abstract domain-independent user goals/tasks with alternative workflows that achieve those goals with various levels of security. Such translation of information is performed with the aid of an AI planner, SHOP2. Consequently, system analysts with no deep experience in security technologies can acquire a view of what steps and technologies are involved in making their designs more secure and implement accordingly.
Open Access
Improving Seafood Production Through Data Science Methods
(2024-11-07) Teimouri Lotfabadi, Bahareh; Khaiter, Peter A.
Global production of seafood has quadrupled over the past 50 years. Seafood production is characterized by one of the highest waste rates in the food industry reaching up to 50% of the original raw material. Therefore, seafood companies are interested in reducing their waste rates, thus increasing production yields. In this thesis, we apply a Data Science (DS) methodology and suggest an extended DS framework to address theoretical and practical issues in the seafood industry. The framework encapsulates data processing, statistical, machine learning, visualization and optimization capabilities. The research employs unique real-world data collected in a seafood production facility over a 2-year period. The study will contribute to the economic well-being of the individual seafood producers as they could perform their business planning and forecasting in a more informed and predictive way as well as to the overall sustainability of the seafood industry due to the waste rate reduction.
Open Access
Integrating Natural Language and Visualizations for Exploring Data on Smartwatch
(2024-07-18) Varadarajan, Kaavya; Prince, Enamul Hoque
Smartwatches are increasingly popular for collecting and exploring personal data, including health, stocks, and weather information. However, the use of micro-visualizations to present such data faces challenges due to limited screen size and interactivity. To address this problem, we propose integrating natural language (voice) with micro-visualizations (charts) to enhance user comprehension and insights. Leveraging a large language model like ChatGPT, we automatically summarize micro-visualizations and combine them with audio narrations and interactive visualizations to aid users in understanding the data. A user study with sixteen participants suggests that the combination of voice and charts results in superior accuracy, preference, and usefulness compared to presenting charts alone. This highlights the efficacy of integrating natural language with visualizations on smartwatches to improve user interaction and data comprehension.
Open Access
Integrating Precipitation Nowcasting in a Deep Learning-Based Flash Flood Prediction Framework and Assessing the Impact of Rainfall Forecasts Uncertainties
(2022-12-14) Mhedhbi, Rim; Erechtchoukova, Marina G.
Flash floods are among the most immediate and destructive natural hazards. To issue warnings on time, various attempts were made to extend the forecast horizon of flash floods prediction models. Particularly, introducing rainfall forecast into process-based hydrological models was found effective. However, integrating precipitation predictions into flash flood data-driven models has not been addressed yet. In this endeavor, we propose a modeling framework that integrates rainfall nowcasts and assesses the impact of rainfall predictions uncertainties on a Deep Learning-based flash flood prediction model. Compared to the Persistence and ARIMA models, the LSTM model provided better rainfall nowcasting performance. Further, we proposed an Encoder-Decoder LSTM-based model architecture for short-term flash flood prediction that supports rainfall forecasts. Computational experiments showed that future rainfall values improved flash floods’ predictability for extended lead times. We also found that rainfall underestimation had a significant adverse effect on the model’s performance compared to rainfall overestimation.
Open Access
Machine learning algorithms for Long COVID effects detection
(2024-03-16) Ahuja, Harit; Litoiu, Marin; Sergio, Lauren
In the realm of the Internet of Things (IoT) and Machine learning (ML), there is a growing demand for applications that can improve healthcare. By integrating sensors, cloud computing and ML we can create a powerful platform that enables insights into healthcare. Building upon these concepts, we propose a novel approach to address the widespread problem of long COVID. We utilize a wearable device to capture electroencephalogram (EEG) readings, which are then transformed through a set of processing steps into actionable decisions. We use a methodology that initiates data collection from a Cognitive-Motor Integration (CMI) task, followed by data preprocessing, feature engineering, and then the application of ML and advanced Deep Learning (DL) algorithms. To address challenges like data scarcity and privacy concerns, we generate synthetic data and train them using the same model as the original data for comparative analysis. Our method was tested on real cases and achieved prominent results: the CNN-LSTM model achieved 83% accuracy with original data and surged to 93% using synthetic data.

Browse

Browsing Information Systems and Technology by Subject "Artificial intelligence"

Results Per Page

Sort Options