https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4 This community include all MCS Postgraduates Theses from 2004 to 2017 2026-03-27T10:25:32Z https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4028 Title: Opinion Mining Restaurant Reviews on Demand Authors: Perera, I.K.C.U. Abstract: Since expansion of social media and internet are driving to a whole another level, most of the users critically review anything on the internet specially foods and services in restaurants to showcase their humble opinion. These opinions are very valuable in decision making process. Analyzing and extracting the actual opinion throughout these reviews manually is practically difficult since there are large numbers of reviews available in the various aspects. So, an automated methodology is needed to solve this problem. Opinion mining or sentiment analysis is such methodology to analysis these reviews and classify topics as positive, negative and neutral. There are three different levels of opinion mining; Document based, Sentence based and Aspect based. Document and Sentence based opinion mining focus on overall polarity of document and sentence respectively and do not describe the important aspects of each opinion which is more accurate. Hence Aspect based opining mining is the trending topic and this thesis is specifically focused on it on reviews in the domain of restaurants. This empirical work is done for restaurant reviews. Aspect extraction and orientation detection has been used to find the output of this research. This proposed system satisfies 70% of the research objective. 2017-01-01T00:00:00Z https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4027 Title: Analyse Near Collision Situations of Ships Using Automatic Identification System Data Set Authors: Chathuranga, W.T. Abstract: Ship-to-ship collisions is an area of interest for many stakeholder groups such as ship owners, regulatory authorities and insurers etc. Ship-to-ship collisions has significant effect on marine causalities. Many researchers have found different methodologies of analysing ship-to-ship collisions. However, analysing actual ship-to-ship collisions is limited by the number of data points available. As a solution, this study utilises the high data availability in near collision situations based on the fact that near collisions are more common than the actual ship-to-ship collisions. This paper studies the near ship-to-ship collision situations using Automatic Identification System (AIS) data set. First part of the study defines a near collision criterion by evaluating existing ship domains available. Evaluation of ship domains includes mathematical ship domains, statistical ship domains and hybrid domains. The selected ship domain is a hybrid model which considers both static and dynamic characteristics of the own ship and the target ship such as Speed Over Ground (SOG), Course Over Ground (COG), width of the ship and the length of the ship etc. In the second part, the study defines the selected ship domain and the near collision criterion with the real time AIS data set extracted from the U.S coastal area. Application of the ship domain and near collision criterion requires the handle large volume of AIS data in this study AIS data of a one coastal zone in a month consisted with approximately 25 million of data records. This study suggests R based methodology to handle the AIS data using the ‘Lazy Evaluation’ concept. And in the final part of the study generates and interpret the descriptive statistics of the identified near collision situations and also evaluates several data mining in detecting the near collisions with the defined criterion. Study summarises the results of the evaluation with the accuracy levels given by each data mining method. Keywords Ship-to-ship collision, Automatic Identification System (AIS), Ship domain, near collision, Speed over Ground (SOG), Course over Ground (COG), lazy evaluation, data mining 2017-01-01T00:00:00Z https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4026 Title: Sinhala Intelligent Word Recognition with Content based Search Suggest Authors: Kahandagamage, K. S. Abstract: Optical Character Recognition is a computer science approach to resolve offline character recognition problem. More advanced approaches like Intelligent Character Recognition and Intelligent Word Recognition are suitable to deal with unconstraint and cursive handwriting. Intelligent Word Recognition approach tries to recognize entire word than individual letters and good approach to process real world documents with unconstraint (free-form), cursive and incomplete handwriting. This research mainly focus on identifying multiline, unconstraint, cursive and incomplete Sinhala words in offline mode with higher accuracy. Identify word lines from scanned image and segment them into primitive components (character or its parts) are considered as prepro- cessing. Image processing methods are used to remove noise, remove frames and underlines, correct skews and slant which increase the accuracy of recognition. Context-free, analytical approach is used to yield a optimum letter string in recognition. Optimum letter string is retrieved by classifying gradient features of a character. 8 directions are considered for feature extraction. Search suggest algorithm with Ayurveda content based corpus is used in post processing to identify words. Natural language processing methods are used to match words by correcting misspelled and incomplete words. Scope of the research is limited to Ayurveda domain but can be extended to any other domain by simply plugging a specific corpus. Prescriptions written by Sinhala PaaramparikaWeda-Mahathwaru are used to validate system accuracy and achieved 62%. 2017-01-01T00:00:00Z https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4025 Title: Train Arrival Time Prediction System for Railway Passengers Authors: Ariyarathne, H. D. S. P. Abstract: The objective of this project is to design an automated system that will track the location of a train and predict the arrival time to a given train station. This automated system will consist of a centralized data analyzing server which will obtain the key information that affect the arrival time, from the passengers already on board the train via a mobile application. The arrival time to all the destinations will be predicted depending on the received Information which is periodically uploaded by the mobile application (i.e. Coordinates of mobile, Speed of the train etc.). It is mainly focused on providing benefits to passengers who travel by train. Aim of this project is to minimize the time wastage of railway passengers and improve the efficiency. It is given the passengers who are waiting at the station an opportunity to select more suitable alternative trains depending on the time of arrival of each train. 2017-01-01T00:00:00Z