http://dlib.iit.ac.lk/xmlui/handle/123456789/1618 2026-04-09T10:32:24Z http://dlib.iit.ac.lk/xmlui/handle/123456789/2189 Identify development disorders that occur in children during their first five years of life using video mining Perera, Binali "Depression is a serious psychological issue that can influence in individuals of all ages, including teenage students and toddlers. It can be difficult to realize that very young children can experience the depression effects. As per the American Academy of Child and Adolescent Psychiatry, around 11% of youthful children experience the effects of depression at any given point in time. Detection is one of our most important defenses to helping those who show the symptoms of depression. While we have many tools at our disposal to detect it in adults and even more mature children and teenagers, there is a gap around detection it in very young children. Through an depth of literature study and in depth of survey it is justified that phycologists have done research in the area of preschool depression and research findings highlighted that identified preschool depression is important to ensure the child future mental health. And identifying depression in these children can be done by detecting their behaviors in day to day life. It is clearly mention in the Child Behavior Check List, the facts to be identified in the depressed preschoolers. Clinical Depression Bio Meter Address the above problem of identifying preschooler depression by analyzing child activities, and predict future depression status of the child and at the same time it suggest treatment plans for parents according to child currant depression status. HMM model approach was used to predict currant depression status according to the given child activities. The HMM was design according to the sign and symptoms related to the CBCL. Time series approach has taken to predict future depression status by analyzing child depression history data. The system accurate to overall 74% of predicting depression in three sub categories, depress, clinical range and normal range. Identifying and treats this mental in the beginning of early age will decrease the number of depression percentage of youth in the world and CDBM will helps to improve world children mental health" 2023-01-01T00:00:00Z http://dlib.iit.ac.lk/xmlui/handle/123456789/2188 Telco Churn Predictor Weerasekera, Namakl "The Telco industry mainly provides prepaid and postpaid mobile packages. Customers select the package based on the information and feedback provided by users. Identifying the package which the particular customer is based on user behaviour and the patterns are continuously changing. Telco marketing officers are keener on introducing postpaid packages to customers as it leads to high revenue and loyal customers. Analysing customer behaviour over a long period of time is challenging and performing the analysis on a large customer base is not feasible. Hence, there is a need for a model that could learn user patterns from historical data and predict a user base with high confidence who are more likely to churn. As prediction is methodical and explainable, the sales officers could promote packages that would not lead to spam for customers." 2023-01-01T00:00:00Z http://dlib.iit.ac.lk/xmlui/handle/123456789/2187 Implementation of Change Data Capture using Apache Hive to improve ETL performance in a Big Data Warehouse Wickramaratne, Malith "A Data Warehouse acts as a centralized repository for millions/ billions of historical data. In order to provide historical intelligence, the data storage platform and the ETL process play vital roles with regards to the performance of a Data Warehouse. Many organizations tend to use Apache Hadoop as the distribution storage platform for large amounts of data, in other words for ‘Big Data’, however Hadoop has its own limitations when it comes to transactional processing such as inserts or updates or deletes. This study aims to improve the performance of these transactions using Apache Hive, and thereby develop a logic to capture only the changed data within the ETL process. The experimented test results show that this method would improve the execution time of Hive queries, hence an improvement in the performance of the overall ETL process, which could result in significant lead time improvements to cater historical intelligence for organizations and its stakeholders." 2023-01-01T00:00:00Z http://dlib.iit.ac.lk/xmlui/handle/123456789/2186 Personality Type Identification System using unstructured text in English based on Natural Language Processing and Machine Learning De Seram, Edirisooriya Mohottige Pamodya Jayangani "Perosnality type identification is beneficial to understand the associates. Especially to understand life partner in order to have a successful marriage life, select the most suitable candidates for a company, and understand and explore the self capacities are some of them. Personality is combination of a person’s behavior, feelings, motivation, and thought patterns. Those characteristics take years to understand and identify in a person’s personality. Personality type identification system was proposed to speed up the process. In the study, the author identified that existing systems have a gap in identifying personality using unstructured text. To provide speedy and accurate information, the author selected to use Natural Language Processing and Machine Learning techniques. Therefore, the author used the Decission Tree algorithm, the K-Neigbour Algorithm, Support Vector Machine, Naive Bayes algorithm, Logistic Regression algorithm, Random Forest algorithm, XGBoost model, and the LightGBM model. Considering the data set analysis, algorithms’ accuracy and evaluation metrics, the author developed Voting classifier ensemble model. To improve the accuracy, user balanced the dataset. If the original dataset was balanced, the author will be able to implement a more accurate model." 2023-01-01T00:00:00Z