Master of Science By Research

Permanent URI for this collectionhttp://192.248.9.226/handle/123/10419

Browse

Recent Submissions

Now showing 1 - 5 of 5
  • Thumbnail Image
    item: Thesis-Abstract
    Business process re-engineering in a logistic operation
    Dissanayake, CS; Cooray, TMJA
    The development of the industrial age has seen a remarkable growth which has led to competition not of products but of the supply chains. A related problem such organisations face is the difficulty in identifying the most appropriate way of managing the operation in a cost effective and efficient for the organisation as a whole. The better way to solve such kinds of problems, is the use of Operations Research (OR) techniques. The purpose of this project is to use statistical techniques to solve operational problems and further optimise the model. Here an operational environment is used to apply this learning with the intension of gaining benefits in terms of cost savings and service improvement. \ Here two operating models (model A and model B) were studied in detail study, its pros and cons as well as problems that may arise were identified. Since the model needed to be cost effective, the main cost elements were identified and their impacts were quantified base on the past information and finally forecast figures were estimated. Based on all the key parameters, the final impacts ofthe models were derived along with the optimum inventory model and the feasibility ofthe model is also evaluated. Finally, the outcomes were evaluated for all the cost elements using the actual data of the two models and the best model has been concluded to be Model B since it is cost effective by 6.5% and also service oriented. At this point, deviations of cost due to inefficiencies in the operation were also identified where the main cause is due to poor inventory management. Therefore, could conclude that proper inventory management is essential in order to optimise model B and for it to be feasible. The inefficiencies were proposed to be solved as future projects. The difficulties faced during the study and limitations are also been discussed. provided such as identifying a better location to relocate an optimum distribution network Finally, recommendations the Regional DC (Distribution Centre) and to develop are to reduce the distribution cost
  • Thumbnail Image
    item: Thesis-Abstract
    Comparing the applicability of box-jenkins arima methodology and arch/garch methodology among real data sets
    (2014-08-19) Ferdinandis, MGSMA; Cooray, TMJA; Malhardeen, MZM
    The main objective of this research is to compare the applicability of Box-Jenkins ARIMA methodology and ARCH/GARCH methodology among two real data sets. This study addresses the question of how to analyze time series data, identify structures, explain behaviours, model the identified structures and using the insight gained, to analyze and forecast values for the specific time series. For the purpose of this study the time series data included, the total kurakkan yield obtained from the Census department of Sri Lanka and the Money series obtained from the International Financial Statistics data source of Central bank of Sri Lanka. Each of the time series has its own characteristics and different methodologies were needed to require a deeper understanding of the time series data. The analysis of time series constitutes an important area of statistics. The kurakkan yield data set consisted of a few missing values. Three different approaches namely deterministic, stochastic and state space method were used to estimate these missing values. Out of the three approaches the state space method gave the best estimates. Once the missing values were fitted The complete series was used to analyze and then forecast values. To build the models and perform the analysis a statistical software called "Minitab" and the software package called "E-views" was used. The best model obtained was a seasonal ARIMA model with 2 non-seasonal AR terms and 2 non-seasonal MA terms with one seasonal differencing. The model was used to forecast values and the accuracy measure MAPE was 1.65% for the ARIMA model fitted, which.was the minimum value of MAPE for all the en-bloc methods mentioned above. The errors of this model were independent and identically distributed and followed a normal distribution. The main difference between the two time series data sets used for this study is that the money series obtained is a high-volatile data series which includes heteroscedasticity. For this data series the ARIMA methodology cannot be used since the data will not become stationary to fit a model. Therefore the ARCH/GARCH methodology was used to deal with the money data series. To build models for this series the software package called "E-views" has been used. Different ARCH models and GARCH models were fitted to this data set and the parameters were chosen so that the kurtosis value was closer to three. The best model was, a logarithmic transformation of the money series with one GARCH term and no ARCH terms. This model yielded a kurtosis value of 3.09. The main model for this data set did not include any AR or MA terms. However a very large number of data points are required to model the series with AR and MA terms in the main model. iii
  • Thumbnail Image
    item: Thesis-Abstract
    Portfolio optimization using quadratic programming
    (2014-08-14) Ranasinghe, LP; Cooray, TMJA; Dissanayake, R
    Investment analysis is concerned, portfolio optimization is very important in order to get maximum profit. In the proposed research the optimization will be done in two main steps. The first part is the modelling mean variance so called reward and risk. The second part is finding optimum solution. The data set published by the Colombo Stock Exchange was used for this research paper as the raw data. The following five companies are selected for the analysis without biases those are Commercial bank, John Keells, Lanka Hospital, The Sri Lanka Telecom and The United motors. These companies represent several fields in the Sri Lankan market such as banking, group of companies, health service, semi government companies, automobile sector. The objective of the research is to find the optimum allocation of the portfolio. The risk should be minimized and the reward should be maximized at the same time. As a strategy to do both of these simultaneously, the linear combination with controlling arbitrary constant is used. That particular linear combination is a convex quadratic function. In order to find the solution of this, the numerical method is used via MATLAB inbuilt'm file'. The developed model of the Markowitz portfolio optimization model1 could be formulated in order to find the optimum allocation of investment amounts for any number of investment channels. The model can be used by investment researchers and could be applied to gain an analytical idea about the efficient frontier. The model has a parameter that can change emphasis on risk minimization or reward maximization. The portfolio optimization finds the optimum allocation of money to be invested. The optimum allocation depends on several factors, according to Markowitz, the return as well as risk, should be considered simultaneously. The main model for this research is 'Markowitz Portfolio Selection Model'. The objective function of the above model consists a linear combination of risk and return. Since the risk is a quadratic expression, the objective function can also be considered as a quadratic function. Then the normal optimization cannot be applied and the non linear optimization (quadratic optimization) must be applied. The main constraint that can be identified is the budgetary constraint along with other limitations, such as boundary restraints. The model has the advantage of changing the budget at any time and the user can use the total budget as a unit, then the optimum allocation fractions, for each investment can be found. The optimization calculation is carried out through 'Matlab', computer aided calculation software. The output of the optimization model is the ratio of the total investment amount to be allocated, the allocated in the percentages of the total portfolio for Commercial Bank, John Keells, Lanka Hospital, Sri Lanka Telecom and United Motors respectively as 0%, 0%, 62%, 38%, and 0%. The minimum function value is - 0.0907, and the function stands for the linear combination of the risk and the reward.
  • Thumbnail Image
    item: Thesis-Full-text
    Calculations on face and vertex regular polyhedra and application to finite element analysis
    (2014-08-07) Jayatilake, UC; De Silva, GTF
    Polyhedron is a solid figure bounded by plane faces. Face and vertex regular polyhedra are the polyhedra whose faces are regular polygons and the arrangement of polygons around each vertex is identical. Here general equations to calculate the properties of the face and vertex regular polyhedra are developed. This includes equations for radius of the escribed sphere and internal solid angle of a vertex. Using these equations the radius of the escribed sphere of face and vertex regular polyhedrda are found including that of Snub Cube and Snub Dodecahedron. It is also shown that sphere is a limiting case of a polyhedron. As application to finite element analysis, approximating the boundary by the sides of the finite elements is proposed. Also a method of defining the Lagrange interpolating polynomial is proposed. 2D tessellations are filling of infinite plane using polygons and 3D tessellations are filling of infinite space using polyhedra. With the piecewise polynomial selected in the above manner it is shown that the only possible regular tessellations that can be used in finite elements are Equilateral Triangle and Square in 2D and Triangular Regular Prism and Cube in 3D. It is shown in general that "any polygon having two axis of symmetry with nodes are selected at vertices cannot be used as a finite element i f its Lagrange polynomial contains the complete polynomial of degree two" and "any polyhedron having a polygonal face with two axis of symmetry and having six or more number of vertices with the nodes are selected at vertices cannot be used as a finite element i f its Lagrange polynomial contains a two variable complete polynomial of degree two".
  • Thumbnail Image
    item: Thesis-Abstract
    Statistical analysis and modeling of factors influencing lung cancer
    Ekanayake, AN; Indralingam, M
    Statistics show that lung cancer occupies the third position among the incidence rates of cancers in Sri Lankan males and this rate is increasing yearly. This research is focused on two main areas. These are to find factors associated with lung cancers and study on time to death after detection of a lung cancer, known as the survival time./ Data collection was done at Cancer Institute, Maharagama (CIM) which is the largest hospital for treatment for the disease in Sri Lanka. Three sources of data have helped in this research study. First one was data in summary format at the CIM. Second was file belongs to each of the patients. Third was the patient's detail form, which is filled by a patient. All together two hundred and sixty two lung cancer patients have come to CIM, in the study period from 1st January to 31st December 2002./ Findings of this research are as follows. Smoking is the main risk factor for lung cancers. People who do occupations in areas uncovered for polluted air have high risk for lung cancer. There is a genetic effect for lung cancer. Consuming alcohol and chewing betel are also considerable factors for lung cancer. Having Tuberculosis is also risk factor for lung cancer. Among four types of lung cancer viz.; Aden carcinoma, Squalors cell carcinoma, Small cell carcinoma and Large cell carcinoma, the most common types in Sri Lanka are Aden carcinoma and Squalors cell carcinoma. Age, sex, religion and smoking habit of the patient have high relationship with those two types of lung cancer. A male person with age greater than 48 years having smoking habit is more susceptible to Squalors cell carcinoma than for Aden carcinoma./ This research shows that the mean survival time of lung cancer patient is approximately 6 months. Treatment given at Cancer Institute, Stage of diagnosis and sex of the patient affect survival time. Treatment mixture reduces risk of death by half compared to single treatment. Our research shows that of a patient is diagnosed for a lung cancer in extended stage, he/she has eleven times more in risk of death than a patient with localized stage. Risk of death for males is three times more than females.