1,721,033 research outputs found
Optimisation in network asset management
No full textThis thesis considers the role of mathematical programming in asset management.Large, extensive distribution networks are the focus of the work. In particular, welook at how to determine the optimal policy for project release. Projects relating tothe replacement of existing assets and network re-design may be prioritized, givencapital rationing and/or performance improvement requirements in a regulatedeconomic environment. We consider the role of two approaches to modelling underuncertainty in determining an optimal policy for project release in network assetmanagement. These are Monte Carlo simulation and fuzzy linear programming. Wefocus on the maintenance and replacement issues of a large distribution network(network structured system) and consider the application of these modellingapproaches to electricity distribution networks. The electricity companies who ownthe UK electricity distribution networks are under pressure to provide a high qualitysupply to customers at a minimum cost.For this particular replacement problem, a zero-one integer linear programmingmodel is proposed for selecting an optimal project portfolio, based on the objectivesand constraints of the network owner. The modelling approach described in thisdoctoral study would extend to the financial investment appraisal of capital projects toa broad range of manufacturing and energy-related industries such as powergeneration, refining in terms of environment issues and water supply.This thesis presents mathematical programming models, using case studies toillustrate some of the appropriate techniques for developing such models. These casestudy models consider uncertainty and use Monte Carlo simulation to generate morerepresentative results. The models demonstrate the usefulness of Monte Carlosimulation from which we can make recommendations about this and alternativeapproaches
Statistical models for match prediction and decision making in sport
No full textIn this study, we investigate models for the prediction of match outcome. Thesemodels are then used to aid decision-making. In particular, we consider battingstrategy in test cricket. This model provides decision support for a team that is aimingto set a target at declaration. We also develop a measure of the importance of a matchin a tournament. Such a measure may be of use in tournament design.Decision-making on the timing of a declaration in test cricket is considered usingmatch outcome probabilities given the state of a game. Logistic regression is used tomodel the effect of covariates, target set and overs remaining, on match outcomeprobabilities. This approach is then extended to establish batting strategy byconsidering run rate and the distribution of runs scored during a partnership. Adecision tool for batting strategy towards a target aimed for is established.The importance of a particular match in a tournament is measured given theoutcomes of all other matches. This method is illustrated for the English Premiership.Match importance is calculated with respect to winning the Championship, relegationfrom Premiership, qualifying for the UEFA Champions League and prize money.Match outcome probabilities for the match of interest are estimated using an ordinallogistic regression model. Covariates that represent the short and long termperformance of the competing teams are used in this prediction model.This thesis makes the following contributions regarding the application of statisticalmethods in sport. A new quantitative approach that considers the optimum declaration"time" in test cricket is developed. We consider this modelling of fundamentalplaying strategy to be novel. We find that a zero-inflated negative binomialdistribution is a good model for the distribution of runs scored in test cricket. Thematch importance measure that we describe extends an existing definition. The matchoutcome model we use for calculating match importance considers novel covariatesrelated to the recent results of teams
Land to the people : peasants and nationalism in the development of land ownership structure in Zimbabwe from pre-colonialism to the Unilateral Declaration of Independence (UDI) period
Get PDFThe space between the Zambezi and Limpopo Rivers now known as Zimbabwe is a diverse state endowed with diverse ethnicities. The vast majority of the people in this space were peasants and cultivators in pre-colonial times. These peasants had a strong attachment to land because of its psycho-spiritual significance as the abode of the ancestors and other natural resources. One of the ethnic groups in this space, the Shona, had a strong attachment to land for cattle which were very important in the Shona traditional religion. The inhabitants of the space Between the Zambezi and Limpopo also traded, specialized in crafts and did small-scale mining. Trade was practiced over a wide area during the Great Zimbabwe period (11th-15th century) with Zimbabwean gold found as far away as China, and Chinese and Syrian goods imported into the country. With the opening of the African continent to overseas trade the peasants took up the cultivation of export crops in exchange for imported goods. The advent of colonialism in the land now called Zimbabwe affected the peasants’ way of life in a big way. Indigenous people suffered extremely as a result of colonial land policy which characterised the transition to western-style capitalism in the country. The British South Africa Company (BSAC), representing international capitalism, carved out large areas of land for themselves thereby affecting the close relationship between land, cattle, traditional religion and the local inhabitants. Land ownership between the colonial administrators and indigenous people created conflict which ultimately stimulated black nationalism in the country. This work therefore examines the relationship between the peasantry and nationalism, and shows how conflict over resources can motivate stronger collective action which may lead the conflict to escalate into an armed national struggle as portrayed by the First (1896-7) and Second (1966-79) Chimurenga (War of liberation) in Zimbabwe
Advances in capital replacement modelling with applications
Get PDFThis thesis addresses the capital replacement modelling problems associated with amixed, or inhomogeneous, fleet and also takes account of the fleet size problem.Applications considered relate to a fleet of buses and a fleet of medical equipment. Theinitial chapters introduce the notion of capital replacement modelling and reviewprevious work in the field, as well as reviewing the fleet size problem. Replacementpolicies are also put in the context of the fleet rather than the context of a 'typical plant'.In the third chapter, we present our first attempt to model capital replacement withvariable fleet size over a finite planning horizon. A two cycle model is developed inwhich the notion of penalty cost for breakdown is introduced. This cost is incurred whendemand is not met. To take account of the cost of unmet demand, a simple failure modelfor plant is proposed. The replacement model is applied to a fleet of ventilators in anintensive care unit of a hospital. In the fourth chapter we develop various models for thecase of replacement of a sub-fleet within a mixed fleet. These models themselves havevariable finite planning horizon of variable length and build on developments describedearlier in the thesis. Other aspects such as the increased cost of sub-optimal policy dueto delayed replacement, smaller replacement sub-fleet etc. are also considered. Themodels developed in chapter 4 are applied, in the following chapter, to a fleet of busesoperated by a Malaysian inter-city bus company. Sensitivity analysis on different factorsis also carried out. Finally the sensitivity of optimal decision policy to the choice of thereplacement model is described in the context of the bus application
Sequential regression techniques with application to the individual sprint in track cycling
No full textThe research work described in this thesis is concerned with processes comprising a sequenceof stages, where states and actions taken during each stage influence the outcome at the end ofthe process. Statistical analysis of such processes using standard approaches can beproblematic due to the potentially large number of covariates that are influential, especiallytowards the end of the process. Therefore, three alternative statistical techniques of increasingcomplexity were developed. These techniques are all based on a sequential approach, inwhich logistic regression models are developed at consecutive stages. These techniques wereapplied to the individual sprint event in track cycling and all successfully gave insight intobeneficial tactics for each stage of the race.The first technique involves considering for each model only covariates related to the currentand previous stages. As such, a sequence of overlapping models is created. This approachsuccessfully enabled stable and easy to interpret models to be created. However, the jointeffect of applying tactics at different stages of the individual sprint could not be determined.The sequential logistic regression technique overcame this limitation by using the score (thelogistic transformation of the probability of outcome) from the model developed at theprevious stage as a covariate in the succeeding model. As such, all prior information can beincorporated into each model. However this score is estimated with uncertainty, which cancause the model parameter estimates to be biased. Furthermore, the effects of this intrinsicmeasurement error were found to propagate through stages, particularly in terms of therelative importance of prior and current states and actions. The novel third technique thereforecombines the sequential logistic regression approach with measurement error techniques toaccount for error in the score
Estimation and forecasting team strength dynamics in football : investigation into structural breaks
Get PDFThis PhD thesis studies the dynamics of team strengths in football. It investigates the presence of structural breaks, which occur when there is a change in parameters that govern dynamics in a time series. In football, such structural breaks occur because of events such as squad changes during transfer markets as well as managerial or ownership changes. Team strengths are estimated across seven seasons of the Premiership and Championship football leagues and then analysed through a time series perspective, based on the double Poisson model with an added dependence parameter for lower scores and an exponential decay factor that adds more weight to more recent matches. This weighting scheme means that a pseudo-likelihood is used to estimate strength parameters. A rolling window approach is used to obtain a time series for the attack and defence strengths of teams in order to investigate the presence of structural breaks. We show that structural breaks are present in the majority of the time series. These present a challenge for the prediction of match outcomes. By not taking parameter discontinuity into account, one is in essence forecasting team strengths for the next match using incorrect parameter values. We then carry out a forecasting exercise. This involves comparing the mean square error of the one-step ahead forecast of team strengths for all teams, using the two most recent seasons as the out-of-sample forecasting period. We find that different models have a smaller mean square error for different teams, but in particular two models stand out as the best ones: a simple random walk and forecasts made by model averaging. Even though the time-varying parameter model performs quite poorly according to the mean square error, it provides the best match predictions for one of our sub-samples. We conclude that different forecasting models that account for structural breaks can certainly improve forecast accuracy, although our findings are consistent with the econometrics literature that no one model forecasts best all the time. Given the prevalence of structural breaks in determining the dynamics of team strengths, this research has important implications for bookmakers and punters in the betting industry to take these matters into consideration when modelling football match outcomes
Capital replacement modelling with a fixed planning horizon
Get PDFFor equipment or plant replacement, when to replace an existing plant, fleet or apart of it, is one of the main concerns in decision-making. The thesis considersthis decision-making problem using capital replacement models with a fixedplanning horizon, and we took at the behaviour of optimal policy in this context.Application of the models is considered and we compare replacement models witha fixed planning horizon with replacement models with a variable planninghorizon models comprising of two cycles. Capital replacement modelling ingeneral and previous work done in the field are reviewed. The main work of thisthesis is the study of the behaviour of optimal replacement policy for a singleequipment/fleet over a fixed planning horizon, with a numerical investigation ofthe behaviour for non-like-with-like replacement. This is extended to describe thebehaviour of optimal policy for replacement of a mixed fleet. A case study ispresented that applies the fixed planning horizon model to a bus fleet; this fleet isoperated by a Malaysian inter-city bus company. Finally we consider thechallenger problem. Throughout, we recommend the use of a fixed planninghorizon model rather than a two cycle variable-horizon model. The rent criterionis also our favoured criterion for decision-making; the rent criterion exists and iswell behaved for all the models described. A dynamic programming approach isimplemented for the like-with-like replacement problem over a fixed planninghorizon for comparison with the economic life modelling approach of this thesis.We discuss the use of the different replacement decision models for supportingreplacement decision-making in practical contexts
Statistical modelling of training and performance using power output and heart rate data collected in the field
Get PDFThis thesis develops statistical models of performance and training that make use ofpower output and heart rate data. These data were collected during training andcompetition, and were recorded every five seconds using a power meter and heart ratemonitor. Using these data, we estimate the parameters of the Banister model of trainingand performance. In principle, knowledge of these parameters allows one to providequantitative decision support for the scheduling of training in advance of a majorcompetition. The methodology proceeds in a number of steps. In the first, measures of both trainingand performance must be specified. The training experienced by an athlete in a singlesession, the training load, can be measured in a number of ways. We use the TRIMPmeasure. This measure in its simplest form is essentially the total number of heart beats ina training session. Then the training loads of successive sessions are accumulated into asingle measure of training up to time t. This we term the accumulated training effect (attime t). Performance during a session at time t is defined as a function of the power outputobserved during the session. We consider various performance measures and describethese in detail in the thesis. Then in the second step, we relate the performance at time t tothe training load up to time t using a regression model, estimating the parameters of theperformance training relationship. The final step is the training optimisation step, wherebythe known training-performance model parameters can be used to specify training loads upto time T that will maximise (in expectation) the performance at time T. We demonstrate the methodology using the training data histories of ten competitivemale cyclists. As each athlete has his own specific characteristics, we should focus onoptimising training and performance individually. We compare and contrast the differentperformance measures that we propose. Our principal findings are that: Banister model parameters can be estimated; that thedifferent performance measures yield different Banister model parameter estimates andtherefore that the performance measure specification is a matter for athlete/coach choice;and that finally the Banister model has a serious shortcoming for the optimisation oftraining. The articulation of this shortcoming is an important contribution of this thesi
Statistical modelling in test cricket
No full textIn this thesis, we focus on decision problems in test cricket. Initially, we address declaration and follow-on decision problems. We then investigate session by sessionbatting and bowling strategy. Later, we extend our analysis to the rating of test cricket players. We also study how the nature and strength of the covariate effects in ourmatch outcome models vary as a match progresses.We model the match outcome given the end of first, second and third innings positions and then use this for decision making. Our declaration models provide a decision support tool to a batting team captain and management to consider the best timing of declarations in the first three innings. Match outcome probabilities (win,draw, loss) are calculated using nominal multinomial logistic regression models. Wealso propose quantitative decision support for batting strategy in the third innings. We approach the statistical problem by supposing that the third innings run-rate and thetarget that the side batting third aims to set its opponent are decision variables. The follow-on decision problem is also briefly considered: should a captain enforce thefollow-on or not? Surprisingly, we find that the decision to enforce the follow-on or otherwise has no effect on the match outcome. We forecast match outcomes in test cricket in play, session by session. Match outcome probabilities are modelled using multinomial regression, with a win, draw, or loss response, and explanatory variables or covariates relating to match state at the start of each session. These probabilities can facilitate a team captain or management to decide on an aggressive or defensive batting strategy for the coming session. These covariates include the lead, wicket resources used, run-rate, a home advantage factor, and surrogates for the state of the pitch (ground effect) and the pre-match strengths of teams. We attempt to compare our results with bookmakers' odds by means of examples.This thesis also investigates how the covariate effects vary from innings to innings and session to session. The nature of the covariates that influence the match outcomechanges as the match progresses. Early in the match, pre-match team strengths have a large effect. This reduces as the match progresses. Home advantage and ground effectappear small and exist only early on.We also extend our analysis to the rating of test cricket players. The rating system is based on player contributions session by session in a test match. This rating systemevaluates the performance of the players taking into account the stage of match in which runs and wickets are earned and conceded and the influence of the runs andwickets earned on the match outcome
A quantitative analysis of sports tournament designs
No full textIn this study, we develop quantitative methodology to investigate optimal designs andfairness in sports tournaments. We propose a number of tournament metrics that can beused to measure the success of sporting contests or tournaments, and describe how thesemetrics may be evaluated for a particular tournament design. Our principal metric isbased on tournament outcome uncertainty. Tournament outcome uncertainty dependson: the structure of the tournament (such as round-robin, knock-out and hybrids ofthese); the relative strengths of the competitors (competitive balance); and theassignment of teams to individual matches or groups (seeding). Tournament outcomeuncertainty is measured using the tournament outcome characteristic which is theprobability that a particular team in the top 100# pre-tournament rank percentileprogresses forward from round R, for all q and R. We show how tournament designs(the structure, seeding policy and progression rule) and competitive balance influencesuncertainty of outcome. We use Monte Carlo simulation to calculate the values of thetournament metrics. Two match prediction models are employed for the simulation ofindividual match outcomes: (1) the double Poisson model and (2) the Bradley-Terrymodel. These individual match outcome models are then used to run a completesimulation of a tournament. The simulation studies show that the tournament design andcompetitive balance have significant effects on the progression of teams in thetournament. Our methodology is illustrated for various tournaments: the UEFAChampions League, the FIFA soccer World Cup Finals and the ICC cricket World Cup.Seeding policy is found to favour stronger competitors, but the degree of favouritismvaries with type of seeding. Reseeding after each round favours the strong to thegreatest extent. A new efficiency measure for a tournament is also proposed based onthe ability of the tournament to discriminate between competitors' strengths (power)and the size of the tournament. Under these notions the round robin is more powerfulbut less efficient than the knockout design. Comparative studies on seeding techniquesshow that random seeding reduces the power and efficiency of the tournament
- …
