http://dlib.iit.ac.lk/xmlui/handle/123456789/863 Tue, 21 Apr 2026 06:24:38 GMT 2026-04-21T06:24:38Z http://dlib.iit.ac.lk/xmlui/handle/123456789/1036 K-means Clustering based ranking system to select best players among domestic cricketers Dissanayake, G.S " Data science is a wide field of study that consist of data systems and processes that aims to use a scientific approach to maintain data sets and derive meaning from data. On the other hand, Machine Learning is the techniques used by data scientists which enables the computers to learn from data. Machine Learning is a part of Data Science. A vast mathematical knowledge and experience is needed when dealing with machine learning projects with complex algorithms. Clustering in Machine Learning is a type of unsupervised learning method. Generally, clustering helps to identify meaningful structures in data sets, generative features and grouping inherent data sets. After clustering data into groups, data points in one group will be different from the others while points in the same group will be similar to other data points. K-means is a very popular and simple unsupervised machine learning algorithm which is used in clustering. This will identify k number of centroids and allocate the data points to the nearest cluster while making sure that the centroids are kept as small as possible. In this research the author was able to come up with a K-means cluster based player ranking system for domestic cricket in Sri Lanka. While there are other systems, they’re not suitable for domestic level. Through this method the author was able to group players according to their strengths using clustering which will be very useful in selection process. This will be hopefully useful to select players into the national team in the future in an unbiased way and expand to school level with other enhancements" Fri, 01 Jan 2021 00:00:00 GMT http://dlib.iit.ac.lk/xmlui/handle/123456789/1036 2021-01-01T00:00:00Z http://dlib.iit.ac.lk/xmlui/handle/123456789/1035 Used vehicle price prediction system Marasinghe, M. A. A. S " With the increase of vehicle imports, used vehicle market has become popular and interesting. With the market popularity and no prior knowledge about used vehicles, many people get fraud to the vehicle dealers and vehicle buyers. Therefore, this becomes an important and interesting problem. In my research, I proposed a system to predict used vehicle price for the Sri Lankan vehicle market. Therefore, I decided to implement a web application that can predict the price of used vehicles. As first priority, I collected the dataset from Ikman.lk, which is a website that can sell and buy used vehicles. I was able to collect 21510 rows and after several phases of data cleaning process there was 17677 rows. In the data preprocessing phase I have used label encoding as well as one hot encoding for the categorical variables. Before selecting an appropriate algorithm for the model, I have used three algorithms and did a comparative study on each other performance. XGBoost algorithm performed better than linear regression and random forest. XGBoost training and testing results were 97.14% and 91.72%. The best performed algorithm optimized with Bayesian optimization and compared the result. Then the web application built with flask using the prediction model which gave the best result." Fri, 01 Jan 2021 00:00:00 GMT http://dlib.iit.ac.lk/xmlui/handle/123456789/1035 2021-01-01T00:00:00Z http://dlib.iit.ac.lk/xmlui/handle/123456789/1034 TWITHERING-TRACKER: Tea withering process time prediction Gamage, R.T " Tea is the second most famous beverage of the world which is widely used. Compared to other crops, tea industry is one of the prominent industries in Sri Lankan Economy. Tea manufacturing has lots of inside processes such as rolling, withering, fermentation and drying. Process of withering and drying mostly depend to the quality and taste of tea. Determining time spent for the entire withering is a huge challenge since it is usually detected by human experiences or sensory details. Considering the technological enhancement in software development, tracking the withering process and estimating the exact time of withering process is an important need for tea industry to reduce cost and time to exceed the expected profits. So, the dissertation focused on designing a solution to address this above problem. The researcher’s intention is to propose a system that can easily track the process of withering from start to end with a machine learning approach. The system was tested with multiple algorithms both in machine learning and deep learning. The proposed system, Twithering-tracker consists with a novel design and program to be worked in industry level. The predictions are generated using very few amounts of input factors which mostly impactable. So, the end users can easily use the system with a minimal effort. Moreover, the proposed system is acted as decision-making system to achieve maximum benefits. The dissertation includes all the findings identified using different approaches. It would be a great contribution since the domain is an untouched area. Furthermore, the dataset used for the implementation can be also taken as a contribution since it was freshly collected by the researcher only for this research purpose. After the final marking phase, the source code, developed model, summary of findings and freshly collected dataset will be published in an open-source medium." Fri, 01 Jan 2021 00:00:00 GMT http://dlib.iit.ac.lk/xmlui/handle/123456789/1034 2021-01-01T00:00:00Z http://dlib.iit.ac.lk/xmlui/handle/123456789/1033 EVOL - AUTONN : Evolutionary automation of neural network Janakan, S " Evol-AutoNN is biomimicry approach to automate the neural network for tabular data using the two biological concepts came from the nature. Biological evolution and the neural network in the human brain are the two-concept used. As the domain of data science keeps on growing, the demands for the tools that make the data science approachable to non-experts will be ever increased. Biomimicry is a new procedure to do the innovation which tries to give strong solutions to human problem by mimicking the patterns and strategies identified in the nature. Triumph of neural network depends on finding the best architecture. An architecture which fundamentally contain the quantity of the neurons in the layer and the activation functions used in them. As the fundamental can be seem as manageable but with scaling up with the workflow makes complicate for both the technical and non technical stakeholders to do manually. Evol-AutoNN is developed to solve the issues mentioned above. It provides novel approach in finding the initial best setting for the user. The user also had the control of selecting the budget which decide the time duration of computational power due to the high concern about computational power. The user has the option in controlling the ranges in the algorithm. EvolAutoNN imitates the process followed in both the natural selection and genetics. It is available as open source which enables everyone to access. " Fri, 01 Jan 2021 00:00:00 GMT http://dlib.iit.ac.lk/xmlui/handle/123456789/1033 2021-01-01T00:00:00Z