ICITR - 2021
Permanent URI for this collectionhttp://192.248.9.226/handle/123/19432
Browse
Recent Submissions
- item: Conference-Full-text
- item: Conference-Full-textPerformance analysis for different optimizers on the cnn model for covid-19 disease prediction based on chest x-ray images(Faculty of Information Technology, University of Moratuwa., 2021-12) Dhanapala, GHGSAD; Sotheeswaran, S; Ganegoda, GU; Mahadewa, KTThis paper analyzes the performance of different optimizers on the Convolutional Neural Network (CNN) model for COVID-19 disease prediction based on Chest X-ray images. The novel coronavirus known as ‘COVID-19’ or ‘Corona Virus Disease 2019’ has become a severe problem for the world community. Reverse Transcription Polymerase Chain Reaction (RT-PCR) can be known as a significant test used to diagnose COVID-19. However, some of the results of these tests were reported as false-negative, and healthcare facilities face limitations on RT-PCR tests since they are costly, complicated, less optimal of sensitivity, and time-consuming. Perceiving these limitations, detection and the classification of COVID-19 using chest X-ray images can be more accurate, faster, and less expensive when considering RT-PCR tests since X-ray imagery is one of the standard methods that has been used for several decades in medical diagnosis. Diagnosis of COVID-19 related to the radiological manifestations using chest X-ray images is unfamiliar since it is a new experience for many experts. The manual investigation is challenging and requires expertise radiologists. Therefore, a more robust 17 layered CNN model is carried out hereafter doing an experimental analysis on five different optimizers such as Stochastic Gradient Descent (SGD), Adaptive Gradient Descent (Adagrad), Adadelta, Root Mean Square Propagation (RMSprop), and Adaptive Moment Estimation (Adam) for the detection of COVID-19 disease based on chest X-ray images. The chest X-ray images under COVID-19 and normal were collected from a multi-class dataset in the Kaggle repository. The proposed model outperformed a training accuracy of 99% and a validation accuracy of 99% with the optimizer Adam along with max-pooling.
- item: Conference-Full-textNovel approach for load balancing in mobile cloud computing(Faculty of Information Technology, University of Moratuwa., 2021-12) Ranapana, RAAIB; Jayasena, KPN; Ganegoda, GU; Mahadewa, KTMobile cloud computing (MCC) was used in many sectors in the current day world, and it combined mobile computing and cloud computing technology to provide mobile services. Mobile devices have short battery life and storage; MCC plays a significant work in compromising that issue. When users are in a hotspot, they face more difficulties and a bad user experience. Load balancing gives a better user experience in the MCC domain. Edge computing and edge clouds are beneficial for load balancing in MCC. There are various algorithms for load balancing in MCC. This research in a simulation environment rather than a natural environment. This research, is focusing on developing a hotspot migration mechanism, reducing mobile devices' battery usage, and developing a novel load balancing algorithm. This research focuses on providing solutions for the limited battery life of mobile devices and the gap in load balancing in mobile cloud computing and provides suggestions to future researchers.
- item: Conference-Full-textAdapting marytts for synthesizing sinhalese speech to communicate with children(Faculty of Information Technology, University of Moratuwa., 2021-12) Lakmal, MAJA; Methmini, KADG; Rupasinghe, DMHM; Hettiarachchi, DI; Piyawardana, V; Senarathna, M; Reyal, S; Pulasinghe, K; Ganegoda, GU; Mahadewa, KTThe majority of the Sri Lankan population speak Sinhala, which is also the country's mother tongue. Sinhala is a difficult language to learn by children aged between 1–6 years when compared to other languages. Text to speech system is popular among children who have difficulties with reading, especially those who struggle with decoding. By presenting the words auditorily, the child can focus on the meaning of words instead of spending all their brainpower trying to sound out the words. In Sri Lanka, however, computer systems based on the Sinhala language especially for children are extremely rare. In this situation having a Sinhala text-to-speech technology for communicating with children is a helpful option. Intelligibility should be considered deeply in this system because this is specific for children. Recordings of a native Sinhalese speaker were used to synthesize a natural-sounding voice, rather than a robotic voice. This paper proposes an approach of implementing a Sinhalese text-to-speech system for communicating with children using unit selection and HMM -based mechanisms in the MaryTTS framework. Although a work in progress, the intermediate findings have been presented.
- item: Conference-Full-textFleet management with real-time data analytics(Faculty of Information Technology, University of Moratuwa., 2021-12) Rathnayaka, RPDT; Ekanayake, KVJP; Rathnayake, HUW; Jayetileke, HR; Ganegoda, GU; Mahadewa, KTSignificant effort has to be devoted to surviving the businesses relying on fleet vehicles in the year 2020 and ahead as the novel coronavirus (COVID-19) epidemic became pandemic. Executing profitable business while keeping the staff safe and productive is a critical challenge to deal with. To find a solution, we focus on driver management out of major functions in fleet management such as vehicle, driver, and operation management. We were unable to identify a study conducted to capture real-time data on a ride in a fleet. Therefore, to fill that gap we implemented a cost-effective real-time Fleet Management System (FMS) using data analytics with the use of ESP32 SIM800L with reprogrammable capabilities. Fleet can use this system to monitor real-time data on vehicle location, remaining time to the destination, vehicle speed, and distance traveled. Moreover, the system can be personalized as it has reprogrammable features to be enabled or disabled based on the customer's preference. Once the data is captured through the Global Positioning System (GPS) receiver, data will be transmitted via General Packet Radio Service (GPRS) to two remote servers. One server is hosted locally with SQL and where the other is hosted in a cloud environment with a Firebase realtime database. The vehicle location is tracked using GPS. For fast data transfer, 3G Global System for Mobile communications (GSM) with ESP32 800L microprocessor was used. A web-based graphical user interface is developed to analyse and present the transmitted data. Vehicle information can be viewed and located on the web application in form of google maps. Real-time data analytics is used with Firebase's real-time database. Furthermore, Short Message Service (SMS) facility is made available for the driver to communicate with configured mobile numbers
- item: Conference-Full-textMachine learning-based automated tool to detect Sinhala hate speech in images(Faculty of Information Technology, University of Moratuwa., 2021-12) Silva, E; Nandathilaka, M; Dalugoda, S; Amarasinghe, T; Ahangama, S; Weerasuriya, GT; Ganegoda, GU; Mahadewa, KTSocial media platforms have emerged rapidly with technological advancements. Facebook, the most widely used social media platform has been the primary reason for the spread of hatred in Sri Lanka in the recent past. When a post with Sinhala hate content is reported on Facebook, it is translated to the English language before the review of the moderators. In most instances, the translated content has a different context compared to the original post. This results in concluding that the reported post does not violate the established policies and guidelines concerning hate content. Hence, an effective approach needs to be in place to address the aforementioned problem. This research project proposes a solution through an automated tool that is capable of detecting hate content presented in Sinhala phrases extracted from Facebook posts/memes. The tool accepts an image that contains Sinhala texts, extracts the text using a Convolutional Neural Network (CNN) model, preprocesses the text using Natural Language Processing (NLP) techniques, analyzes the preprocessed text to identify hate intensity level and finally classifies the text into four main domains named Political, Race, Religion and Gender using a text classification model.
- item: Conference-Full-textHeuristics-based sql query generation engine(Faculty of Information Technology, University of Moratuwa., 2021-12) Sugandhika, C; Ahangama, S; Ganegoda, GU; Mahadewa, KTA database is one of the prime media to store data. Most of the time, relational databases are preferred over other databases due to their ability to represent complex relationships between data. Languages like Structured Query Language (SQL) are used to retrieve data stored in relational databases. Information stored in these databases is often accessed by naïve users who do not possess high competencies in technical database querying. Therefore, Natural Language Interfaces to Databases (NLIDB) are being developed to translate natural language into SQL queries and retrieve the corresponding database results. This paper proposes a novel NLIDB called SQL Query Generation Engine which has been developed using a heuristics-based approach. The system was tested with more than 200 natural language queries and has shown an overall accuracy of 93%.
- item: Conference-Full-textCheating detection in browser-based online exams through eye gaze tracking(Faculty of Information Technology, University of Moratuwa., 2021-12) Dilini, N; Senaratne, A; Yasarathna, T; Warnajith, N; Seneviratne, L; Ganegoda, GU; Mahadewa, KTEye-tracking can detect and examine human visual attention, emotional conditions, latent cognitive processes such as efforts to recall a concept or the fear of running out of time, and so on. Hence, we can use eye-tracking to identify deviant behavior patterns in learning and problem-solving. At present, given the existence of a global pandemic, online exams are widely used by educational institutions to evaluate students' performance. However, identifying cheating is challenging due to the absence of a human (invigilator) monitoring students' behavior as done in exams held in a physical location. In an online environment, students' behavior, and attempts to cheat, can only be captured via a computer, thus requiring a mechanism for online proctoring with capabilities for cheating detection. In this research paper, we present a browser based cheating detection approach in online examinations through eye gaze tracking. We developed a browser plugin to track the eye gaze movements through the in-built web camera. Using the plugin, we generate an eye gaze dataset while a student faces an online examination. We then process and analyze this dataset to detect any misbehavior during an online examination. The underlying research work of this paper identifies different eye gaze patterns during online examinations and present a cheating detection mechanism. For anomaly detection in the eye gaze data, we use a One-Class Support Vector Machine (OCSVM). We then use these identified anomalies to predict cheating behaviors of the test takers. The given approach can be used for any web-based quiz examination such as academic institutions' exams, company recruitment exams, and overseas testing exams to detect any anomalous behaviors of the test takers during the examination period. The given eye tracking approach can also be applied to other research domains such as online gaming, and web usability studies to capture information related to user behaviors.
- item: Conference-Full-textOntology based fake news detection for Sinhala language(Faculty of Information Technology, University of Moratuwa., 2021-12) Bandara, O; Jayarathne, D; Shashinika, D; Ranathunga, L; Ganegoda, GU; Mahadewa, KTWith the current trend in extensive use of internet technologies, people are accustomed to a life of being online. Similar phenomena apply for news awareness as well where social media has become one of the main sources of information of these tech-savvy people. Even though it gives easy access to information, there is a problem with the trustworthiness and authenticity of the news posted on social media. As such, fake news detection in social media platforms has become an active research area. Despite that, for the Sinhala language, there are only a few attempts carried out to detect fake news. With this background, this research project has come up with a novel idea of ontology-based fake news detection for news published using Sinhala language. We believe that content-based fake news identification is the most appropriate method to assess the truthiness of a news article and our system was able to give promising results in detecting fake news.
- item: Conference-Full-textDiagnostic intervention for mental disorder(Faculty of Information Technology, University of Moratuwa., 2021-12) Senanayake, S; Karunanayaka, C; Dananjaya, L; Chamodya, L; Kumari, S; Chandrasiri, S; Ganegoda, GU; Mahadewa, KTMental health is one of the essential factors in the topic of healthcare and wellbeing. However, mental health disorders could cause severe damage, even loss of life to the person or the surroundings, if mental health disorders were not identified and appropriately cured. Unfortunately, though there is good help there, some people have a hard time detecting whether they are suffering from mental health disorders or not. In this study, the team proposes a system to detect mental health issues using facial emotion recognition (FER), sleeping patterns, social media web scraping, and heart rate. The intention is to give an accurate prediction of the mental health status of a person using these three nodes.
- item: Conference-Full-textA rule based approach for hemorrhage detection in digital fundus photographs(Faculty of Information Technology, University of Moratuwa., 2021-12) Munasingha, SC; Pathirana, P; Priyankara, KK; Upasena, RG; Ikeda, A; Ganegoda, GU; Mahadewa, KTHemorrhages are one of the earliest signs of Diabetic Retinopathy, hence accurate detection of hemorrhages is crucial in an automated DR detection system. In this paper, a novel and robust rule based methodology for automated detection of hemorrhages is proposed. We present an ensemble technique for hemorrhage classification by incorporating size-based classification, color-statistic-based classification, and shape-based classification along with a novel dual step filtering approach for candidate detection. Finally, we present an experimental study carried out on DIARETDB database using the proposed method to detect and segment hemorrhages in retinal images.
- item: Conference-Full-textComputational modelling of synaptic plasticity: a review of models, parameter estimation using deep learning, and stochasticity(Faculty of Information Technology, University of Moratuwa., 2021) Kumarapathirana, KPSD; Kulasiri, D; Samarasinghe, S; Liang, J; Ganegoda, GU; Mahadewa, KTIt is imperative to understand the human memory formation and impairment to treat dementia effectively. There is ample scientific evidence that memory formation is strongly correlated to synaptic connections. Synaptic plasticity reflects the strength of these connections and is strongly related to memory formation and impairment. The complexity in the signalling pathways and interactions among proteins demands a systemic approach to study synaptic plasticity. Hence systems biology approaches are used in computational neuroscience. In this paper, we review the key computational models related to synaptic plasticity, the use of deep learning in parameter estimation, and the incorporation of epistemic stochasticity in the models.
- item: Conference-Full-textDetermining flood risk vulnerability using factor analysis approach(Faculty of Information Technology, University of Moratuwa., 2021-12) Karunarathne, AWSP; Piyatilake, ITS; Ganegoda, GU; Mahadewa, KTFloods are becoming a frequent natural disaster in Sri Lanka. Although it is an uncontrollable natural event, it is essential to pay attention in necessary policy making in order to control floods in the future. Therefore, this study targets to develop a framework to identify the high risk areas in Sri Lanka. This provides useful information to the government as well as the policymakers to take necessary preparedness, prevention, and awareness actions to lessen the risk from floods. This study uses principal component analysis (PCA) to explore the flood situation in Sri Lanka and to rank the main regions according to the risk level by considering flood risk and controlling factors. Twelve factors including both flood risk and controlling indicators are identified based on prior literatures and then the original data set is constructed. With the aid of the PCA method, four principal components are identified, and they are human induced factor, natural induced factor, human induced controlling factor, and natural induced controlling factor. Based on the weights of the principal components a comprehensive score is derived. Finally, the main regions are ranked using the comprehensive scores. The results reveal that Rathnapura, Kalutara, Colombo, Kurunegala and Matara regions have a high-risk chance of having floods.
- item: Conference-Full-textTaxonomic identification of sri lankan freshwater fish based on advanced feature extraction techniques(Faculty of Information Technology, University of Moratuwa., 2021-12) Semapala, GDCH; Sandanayake, TC; Ganegoda, GU; Mahadewa, KTSri Lanka is a tropical composed of different kinds of animals living in different environments. Among these, various kinds of fish species can be identified around Sri Lankan rivers and basins. Freshwater fish species vary between marine and brackish forms. Some human activities destroy their environment and, as a result, Sri Lankan freshwater fish species are at risk. Consequently, the implementation of the freshwater fish classification system has become very important to remedy this situation. Research indicates that Malpulutta Kretseri, Belontia Signata, and Puntis Tittaya are the freshwater fish species selected for classification. The main objective is to extract features more precisely and accurately while optimizing each feature extraction technique to the optimum level. Initially, four algorithms were used, and checked the results were. Then, two better-performing algorithms namely, SIFT and ORB were sorted out and carried out further tests. These two algorithms used corners, blobs, and edges to extract features. Furthermore, the test was done by segmenting as Body, Head, and Fins, and the results were improved significantly. For implementing the system 1000 data training images and 180 data testing images and data validation images were used. ORB algorithm gives 96.7% accuracy and SIFT algorithm gives 85% accuracy. The segmentation method adapted to the characteristics results in a precision of 82%. According to the research, the ORB algorithm-based feature extraction is the more sophisticated technique.
- item: Conference-Full-textDetection of suicide ideation in twitter using ann(Faculty of Information Technology, University of Moratuwa., 2021-12) Yatapala, KDYHT; Kumara, BTGS; Ganegoda, GU; Mahadewa, KTSuicide is considered one of the leading problems in the present. Detecting suicide earlier and providing a solution is considered the most successful way to suicide ideation and suicide attempts prevention. At present, online communication channels are used to express the suicidal tendencies of some people. This paper presents a machine learning approach to identify suicide pattern and detect suicide ideation or thoughts by considering online user-generated content with the aim of suicide ideation detection. People who have suicidal ideations, express strong negative feelings. Here, an Artificial Neural Network is used as a machine learning algorithm. To detect ideations of suicide, we generate feature vectors using different techniques including Word2Vec, Doc2Vec, and TF-IDF features. As the online user communication channel, we select Twitter.
- item: Conference-Full-textA framework to detect sale forecasting with optimum batch size(Faculty of Information Technology, University of Moratuwa., 2021-12) Saradha, RMS; Samadhi, MA; Manawadu, I; Ganegoda, GU; Ganegoda, GU; Mahadewa, KTToday, sales forecasting plays a key role for each business. To maintain the sales process successfully, every manufacture focus on retaining optimum production batch size. Therefore, this study aims to develop a framework to detect sale forecasting with optimum batch size. This work focuses on predict future sales and optimum production batch size by using different machine learning techniques and trying to determine the best algorithm suited to the problem. Here, Auto-Regressive Integrated Moving Average (ARIMA) model is used to predict future sales and Artificial Neural Network (ANN) model is developed to determine the optimum level of production as a function of product unit, setup cost, and holding cost in our approach and have found these models have better result than other machine learning models.
- item: Conference-Full-textUsage of topic modeling method for high dimensional gene expression data analysis(Faculty of Information Technology, University of Moratuwa., 2021-12) Senadheera, SPBM; Weerasinghe, AR; Ganegoda, GU; Mahadewa, KTGene expression data analysis is a major area in biological system interpretation. Since, gene expression data have large numbers of variables, high dimensional clustering methods are required for analysis. The objectives of this study were to understand the effectiveness of different clustering methods in gene expression data analysis based on biological relatedness and study of the advantages and disadvantages of different clustering strategies in gene expression analysis. The data was obtained from the GSE19830 dataset and the brain tumor data (TCGA project). To test the hard clustering, hierarchical clustering and fuzzy clustering, the K-means algorithm, HClust and topic modeling were used respectively. Prior knowledge about the dataset was required to define the number of clusters (K). Initially, the GSE19830 (Brain, Lung, Liver tissue mixture) dataset was used for developing the clusters. All models clustered the observations similar to the physical tags in the dataset. Secondly, Clustering methods were developed with the brain tumor dataset consisting of 202 samples (four specified physically categorized tumors). According to hierarchical clustering and topic modeling, when analyzing similar tissues, gene expression tumor subtypes (clusters) were not aligned with physical categorization. Finally, 81 cancer genes were filtered and generated a topic model. In order to understand the biological relevance of the final model, Reactome and PCViz tools were used. Reactome results supported topics developed from topic modeling. According to the results, in high dimensional data analysis, topic modeling was found to be a promising approach for gene expression based clustering while K-means was found to be inappropriate for gene clustering.
- item: Conference-Full-textDigital platform to empower the self-employment in Sri Lanka(Faculty of Information Technology, University of Moratuwa., 2021-12) Wickramasinghe, HCP; Thebuwana, TD; Wijesinghe, GKHS; Dissanayake, UN; Kodagoda, N; Suriyawansa, K; Ganegoda, GU; Mahadewa, KTUnemployment is a huge problem around the world because a lack of job opportunities. People are unable to find the job opportunities according to their preferences and qualifications. As a solution for this, many countries are attempting to empower self-employment. Most of current world problems have been solved using modern technologies. Therefore, the development of self-employment also can be achieved through modern technology. The objective of our proposed platform, HIRELANCER, is empowering self-employment using modern technologies. HIRELANCER is bringing the consumers, service providers, and suppliers into the same platform. HIRELANCER will consist of innovative features that go beyond comparatively to other platforms such as an advanced mechanism to find best suitable service providers/suppliers for the service, handling the virtual front-desk, cost estimation for the services prior to contacting a service provider, and advanced facility to find a suitable career path for the people who are seeking career guidance. This research paper discusses how the innovative features of HIRELANCER will be beneficial for consumers, service providers, and suppliers and ultimately achieve our main objective, which is empowering self-employment in Sri Lanka.
- item: Conference-Full-textA data driven approach for detection and correction of spelling errors in sinhala essays(Faculty of Information Technology, University of Moratuwa., 2021-12) Samarasinghe, PM; Sewwandi, WBI; Ranathunga, L; Wijetunge, WASN; Ganegoda, GU; Mahadewa, KTThis paper proposes novel approaches for checking and correcting spelling errors in Sinhala essays written by candidates of grade five scholarship examination. They don't have a proper mechanism to identify their spelling mistakes in essays by themselves. Spelling errors by such students may occur due to the violation of spelling rules, missing or adding of letters, missing modifiers, inaccurate spelling in a similar structure, and similar sound letters]. To mitigate such challenges, the Sinhala corpus file has been developed to identify the accurate and inaccurate spellings of the written words. The role of this application is to identify the correct and incorrect words which are entered by the user and generate the most correct words as suggestions for the incorrect words. This paper introduces three new novel approaches to detect the correctly spelled words in Sinhala essays namely object word checker method, suffixes checker method and similar word checker method. With addition to that this paper discusses three approaches to generate accurate suggestions including one novel approach. When evaluating the accuracy of the spelling error detection and correction module the overall results for precision, recall, and the f -measure were recorded as 83.05%, 85.57%, and 86.62% respectively.
- item: Conference-Full-textGis powered an automated generic flood model for river basins in Sri Lanka(Faculty of Information Technology, University of Moratuwa., 2021-12) Bandara, MNL; Premasiri, HMR; Sudantha, BH; Ganegoda, GU; Mahadewa, KTFlood is the most common and deadliest form of disaster that affects lives and properties all around the world. Predicting natural disasters is very complex due to the lack of proper methods and resources in countries like Sri Lanka. But if there is an efficient prediction system it helps to save not only lives but the environment and infrastructure too. Therefore, the aim of this study is to pave the pathway to build an efficient and effective flood prediction system through analysing available flood modelling techniques and their applications to find their strengths and weaknesses. Then the result of the study could be used to put the foundation for the main requirement of building the system to predict natural disasters. A generic model was developed to take any DEM data and a pour point feature class layer for the specific DEM to generate outputs based on other variables that could be input to the model. It gave model calibration capability as well as significant time saving on tasks. The use of special tools like the ‘Parse Path’ tool in ArcGIS Pro, gave the capability to name output easily and quickly. And it also made saving so efficient because it automatically saves all the results to the file path of the DEM. Due to these factors, when it starts raining in the upper catchment area, could forecast due inundation area in minutes. Including the GIS technology could improve the data quality and availability while incorporating different data sources for more in-depth analysis could give more accurate predictions. Using the GIS-based hydrological model, a suitable system to implement in Sri Lanka could be developed.