Research Papers - Dept of Information Technology
Permanent URI for this collectionhttps://rda.sliit.lk/handle/123456789/593
Browse
15 results
Search Results
Publication Embargo Image Processing-Based Solution to Repel Crop-Damaging Wild Animals(Springer, 2023-02-03) Fernando, W. P. S.; Madhubhashana, I. K.; Gunasekara, D. N. B. A.; Gogerly, Y. D.; Karunasena, A; Supunya, RTwo-thirds of Sri Lanka’s population is directly dependent on agriculture, which generates one-third of the nation’s GDP. However, crop efficiency in Sri Lanka has declined over the years due to several issues including sub-farm maintenance, destruction caused by wild animals, and unethical farming practices. Among them, the destruction caused by wild animals has led to conflicts between animals and humans causing loss of both animals and human lives in the past. There are a number of technical solutions proposed to solve the above problem, especially in the form of animal repellants. However, such solutions have several limitations, such as the small number of animal groups to be identified and the short distances they can be detected, and the lack of understanding of harmful animal populations. This research proposes an animal-repellent methodology considering several features of animals such as colors, coats, shape, and noise made by animals both in daytime and nighttime. The number of animals approaching crops is also detected and the behavior of animals is monitored to avoid false alarms. The research uses a wide range of techniques such as image processing and deep learning for the above purpose on audio, visual, and image data sets collected from the mentioned animal groups. The solution demonstrated a 90% accuracy for animal identification during the day, and 84% accuracy for animal 2 W. P. S. Fernando et al. identification at night, whereas the accuracy of studying animal behavior patterns is 90% and animal sounds were identified with 87% accuracyPublication Embargo Comparison of ARIMA and LSTM in Forecasting the Retail Prices of Vegetables in Colombo, Sri Lanka(IEEE, 2022-12-09) Fonseka, D.D; Karunasena, AIdentification of vegetable price trends is important to make better decisions in the production and market. Due to several factors, including seasonality, perishability, an imbalanced supply-demand market, customer choice, and the availability of raw materials, vegetable prices fluctuate quickly and are highly unstable. In this study price prediction was concluded using two models ARIMA and LSTM with retail price data for Cabbage, Carrot, and Green beans in Colombo from 2009 to 2018. According to the decision criteria of RMSE and MAPE, the LSTM model is superior to the ARIMA model in predicting the retail prices of vegetables. There were no studies have focused on predicting prices with novel technology in the Sri Lankan vegetable market. Hence the results of this study can be used to build an advanced forecasting model by the government and decision-makers in agriculture in Sri Lanka.Publication Embargo Machine Learning Based Solution for Improving the Efficiency of Sugar Production in Sri Lanka(IEEE, 2022-12-26) Kulasekara, S; Kumarasiri, K; Sirimanna, T; Dissanayake, D; Karunasena, A; Pemadasa, NAlthough sugar is a popularly used commodity in Sri Lanka, sugar manufactured within the country fulfill only a very small portion of the demanded amount. Sugar production is an intricate process which requires a considerable amount of expertise especially in the areas of cultivation, production and revenue prediction which may not exist in novice farmers. This research proposes a methodology which provides novice sugarcane farmers with expert knowledge on four main areas related to farming including weather forecast, sugarcane maturity estimation, production forecast and prediction of return sugarcane amounts from lands. ARIMA model is used for weather forecast whereas machine learning methods and multiple regression models were used for sugarcane maturity estimation and production of forecasts and returns respectively. The final ARIMA time series model was validated with p-value greater than 0.05 for Ljung-Box test with three different lag values. The Support Vector Machines model was identified as the best model with an accuracy of 81.19% for the sugarcane maturity estimation. The SVM model was trained using the HSV and texture features extracted from sugarcane stalk images using image processing techniques. The prediction of sugar production received a testing R-squared score of 87.75% and mean squared error of 0. Prediction of yield received a mean squared error of approximately 0 and R squared score of 98% on test data. The methodology used in this research could be used by novice farmers to increase their cultivation as well as sugar production.Publication Embargo A Meta-learning approach to Predict Non-performing Loans in Sri Lankan Financial Institutions(IEEE, 2022-12-26) Kavirathne, G. P. R. A.; Perera, V. A. S.; Karunathunge, L. C. R.; Dewapura, B. N.; Karunasena, A; Pemadasa, M. G. N. M.Most financial institutions make the majority of their income from loan interest. However, due to the current financial crisis in Sri Lanka, non-performing loans have been the focus of financial industry concerns. Before the crisis, financial institutions were more ready to lend to businesses and individuals. Since the crisis, the rate of non-performing loans has increased, limiting the company’s growth. Predicting the likelihood of nonperforming loans can help in lowering this credit risk. Therefore, this paper presents a machine learning approach to predict nonperforming loans of a financial company in Sri Lanka. Moreover, this study also attempts to develop a meta-model that combines various classifiers, including K-Nearest Neighbors, Decision Tree, Random Forest, Support Vector Machine, and Naïve Bayes. The meta-model is also compared with the baseline models. The predictive performance of all models is compared using accuracy, precision, recall, F1-score, and AUROC score. At the end of the study, it was identified that the meta-learning model is the most effective model to handle this case, with a classification accuracy of 93.09%, precision of 82.68%, recall of 92.17%, F1 score of 86.24%, AUROC score of 96.0%, sensitivity of 92.44% and specificity of 93.17%.Publication Embargo A Multimodal Interviewee Evaluation Approach for Candidates Facing Video Interviews(IEEE, 2022-12-26) Maddumage, T.A. R; Liyanage, K.L.O.G.; Karunasena, A; Weerasinghe, K.M.L.P.; Yasiru Randika, W.G.Automated video interview evaluation is an area that is becoming trending because of the significant usage of video interviews for candidates’ hiring processes. Though there is much research on video interviews, there are a smaller number of studies on the automated evaluation process in video interviews. Still, many entities use more human interaction for selecting candidates through video interviews. This study proposes a set of methodologies to evaluate the candidate by considering five aspects which are knowledge level, mindset, confidence, personality, behavior, and English language fluency. This paper proposes an answer evaluation using feature engineering and Siamese LSTM architecture where BERT is used as the encoding layer. The CNN-based approach is proposed for behavior and personality analysis. For language fluency and confidence, NLP techniques and machine learning methods are proposed in this study. Through the methodologies propose in this study, overall, all the evaluations yield around 85% - 90% accuracy. The approach suggested in this study will help organizations to do their talent acquisition process more smoothly.Publication Embargo Automated Analysis of Children Emotion Expression Levels(IEEE, 2022-08-25) Nadeeshani, N; Kalaichelvan, K; Karunasena, A; Samarasinghe, PDespite the advancement in the field of facial emotion expression analysis, less attention has been given for facial emotion expression and emotion level analysis in children. This paper presents three novel findings in the area of child emotion expression. Identifying and validating the AU stimulation of children, automating the child emotion and level of emotion prediction and age wise analysis of child emotion expression. Emotion predictions were compared resulting through deep learning methods such as 3DCNN and machine learning approaches using EFA.AU stimulation results generated through EFA are consistent with the FACS. Through AU analysis, the paper shows that a child video or image can be predicted for the expressed emotion and its level with 91.04% accuracy through KNN classifier. While the 3DCNN approach resulted in 82.64% accuracy, the age wise emotion prediction through CNN resulted in the range of 60% to 86.6%. Though all approaches evidenced comparable results in emotion prediction, the emotion level prediction through EFA and AU outperformed 3DCNN and CNN approaches in all cases. Happy emotion prediction in age wise emotion analysis resulted in a higher accuracy over sad and disgust emotions. As emotion level prediction in age wise analysis display mixed results, a further research on age wise AU stimulation is encouraged.Publication Embargo Know More: Social Media based Student Centric E-learning platform with Machine Learning Approaches(IEEE, 2022-07-18) Malavige, O; Nasome, V; Costa, M; Jayasinghe, B; Karunasena, A; Samarakoon, USocial media has become increasingly popular among the younger generation in the last decade. Students engage with social media on daily basis, and it affects their interests, lifestyle, and attitude. There are many existing e-learning applications used by higher educational institutes, but such applications are mainly focused on delivering teaching content rather than facilitating active and interactive learning. This paper proposes a novel e-learning platform to create an active and interactive learning environment for students leveraging social media strategies, especially those of “Facebook.” The objective of this platform is to promote self-motivation, self-learning, and interaction. The platform features were built on considering three aspects important for learning, which are personal knowledge management, learning management, and collaborative learning. Features of the proposed platform that it comprises are Newsfeed, Classmates, Profile, Cluster, Repository, Knowledgebase, Bookmark, Topic Map, Search Engine, Test Mark Prediction, and Slide Show Summary generator. Machine Learning techniques and Natural Language Processing were used to build some of the platform features. The feedback collected on the proposed system, “KnowMore,” shows that the satisfaction of the students has increased with the system.Publication Embargo eVision - A technological solution to assist vision impaired in self-navigation(IEEE, 2022-07-18) Hewagama, K. G; Suwandarachchi, T. D; Hettiarachchi, C. R; Alwis, P. L. D. N; Karunasena, A; Weerasinghe, K. M. L. PVisually impaired people face many difficulties in navigation such as crossing the road, identifying signs and text in indoor and outdoor environments and avoiding obstacles. Even though much research has been done to assist visually impaired people, most methods are unpopular, and almost all visually impaired people still only rely on the white cane. This paper proposes eVision which consists of a mobile app as well as a wearable tool that enables visually impaired people to detect obstacles and objects such as moving vehicles and staircases, identify signs, provide assistance with road crossing and natural scene text recognition, using Convolutional Neural Networks and image processing techniques. The CNN architecture used for object detection was SSD Mobilenet V2, since it provided around 95% accuracy for most objects with good performance on mobile. Mobilenet V2 transfer learning model was used for classification of objects, which provided around 94% accuracy. For text detection, the EAST algorithm was used, and the method resulted an accuracy around 98%. From the generated data from models, eVision provides audio feedback to the user using a text to speech(TTS) system.Publication Open Access Machine learning approach for predicting career suitability, career progression and attrition of IT graduates(IEEE, 2021-12-02) Bannaka, B. M. D. E; Dhanasekara, D. M. H. S. G; Sheena, M. K; Karunasena, A; Pemadasa, NThe IT industry in Sri Lanka is associated with a massive work force consisting of skillful professionals and it also provides many job opportunities for fresh graduates at the present. For a fresh graduate entering the IT industry there is a wide variety of job opportunities available and in order to have a satisfactory and rewarding career they should identify the most suitable career for them. On the other hand, employees change their careers and regularly seeking for career advancements and more benefits while the employers struggle to retain employees. Under such circumstances, this research focuses on developing a career mentoring system which comprises of the prediction of career suitability, career and salary progression, and employee attrition to assist IT employees to achieve career goals by overcoming barriers in their career path. For this purpose, data are collected from IT employees, and several models were implemented using classification algorithms such as XGBoost, Random Forest, Support Vector Machine, K-Nearest Neighbors, Decision tree, Naive Bayes, and their performance are compared using accuracy, precision, recall, and F1-Score to select accurate models. XGBoost resulted with higher accuracies for prediction of career suitability, initial salary, career and salary progression with values of 92.31, 90.35, 86.45 and 88.76 respectively. Furthermore, for the prediction of professional courses and employee attrition, Random Forest resulted higher accuracies of 93.52 and 89.70. The ultimate goal of this research is to guide IT graduates and employees to have better performances and to assist them in embracing responsibilities throughout their career life.Publication Embargo Machine Learning Based Emotion Level Assessment(IEEE, 2021-12-09) Wickremesinghe, L; Madanayake, D; Karunasena, A; Samarasinghe, PWith recent advancements of technology, identification of emotions of humans via facial recognition is done with the application of numerous methods including machine learning and deep learning. In this paper, machine learning techniques are applied for identifying different levels of emotions of individuals for unannotated video clips using Facial Action Coding System. In order to archive the above, first, two methods were experimented to obtain a labeled image data set to train classification models where in the first method, clustering of images was done using Action Units(AU) identified from literature and the emotion levels of the images were determined through the resulted clusters and images are labeled according to the cluster they belonged to. In the second method, the image set is analyzed explicitly to identify AUs contributing to emotions rather than relying on those identified in literature and then the clustering of image set was done using those identified AUs to label the images similar to the first method. The two labeled data sets were used to train classification models with Random Forest, Support Vector Machine and K-Nearest Neighbour algorithms separately.Classification models showed better accuracy with data set produced using the second method. An overall F1 score, accuracy, precision and recall of 87% was obtained for the best classification model which is developed using the Random Forest algorithm to identify levels of emotions. Identifying the AU combinations related to emotions and developing a classification model for identifying levels of emotions are the major contributions of this paper. The results of this research would be especially useful to identify levels of emotions of individuals who are having issues in verbal communication.