[dipl] Perustieteiden korkeakoulu / SCI

Permanent URI for this collectionhttps://aaltodoc.aalto.fi/handle/123456789/21

Browse

Browsing [dipl] Perustieteiden korkeakoulu / SCI by Department "Matematiikan ja systeemianalyysin laitos"

Filter results by typing the first few letters
Now showing 1 - 20 of 58
  • Saavutettavuuspohjainen matkojen suuntautumisen ja kulkutavan valinnan simulointimalli
    (2011) Salomaa, Osmo
     School of Science | Master's thesis
    Transportation models are needed to assess the impact of planned transportation and land use projects and operations. Models currently in use are to a large extent outdated and theoretically problematic and are poorly applicable to many current problems. The goal of this work is to develop a new kind of transportation model, one that takes trip chains into account, includes a time-dimension and in which ordinary daily trips and related decision-making is handled by simulation. The model is implemented for the Helsinki commuting area using the spatial precision of a regular 250 meter grid. The main purpose of the model is to investigate how changes in land use and transportation system affect traveling. Results from the model are examined for the present day and for the forecasted year 2035 according to a draft of the 2011 Helsinki region transport system plan. According to the results, the model looks promising and its strengths and limitations have been identified for further development and use. The model is applicable in the short term especially for applications that require spatial precision or inclusion of pedestrian and bicycle trips in analysis and in the long term it can be considered a possible candidate to replace traditional aggregate models.
  • Adaptive Emotion Based Decision Model for a Social Robot
    (2012) Sutinen, Martti
     Perustieteiden korkeakoulu | Master's thesis
    This thesis introduces a computational model of emotions and decisions for a robot, which interacts meaningfully in a social context. The decision making framework is based on multi-attribute utility theory, but it contains a dynamic and adaptive emotional model which basically acts as a preference and perception manipulator. The emotional model is based on event appraisal with discrete emotion categories. Events are assessed using dimensions of utility and probability as well as expectations. The model uses the concepts of core affect and attributed affect to create a multilevel emotion consisting of moods and emotional events. Personality traits are used to create different emotional dynamics by modifying relevant parameters. Attitudes and relationships, understood through attributed affect and classical conditioning, make the robot emotions more believable. The robot learns from user actions and makes predictions about them and environment changes according to probabilistic models. Subjective well-being and human need hierarchy are used as the basis for the preferences which the visceral state affects. The model is inspired by the computational models Cathexis, FLAME, EMA, TAME and Roboceptionist, and is an expanded version of the model used in AISoy1 robot. The framework combines extensive psychological research and requires validation
  • The algebraic square peg problem
    (2014) van Heijst, Wouter
     Perustieteiden korkeakoulu | Master's thesis
    The square peg problem asks whether every continuous curve in the plane that starts and ends at the same point without self-intersecting contains four distinct corners of some square. Toeplitz conjectured in 1911 that this is indeed the case. Hundred years later we only have partial results for curves with additional smoothness properties. The contribution of this thesis is an algebraic variant of the square peg problem. By casting the set of squares inscribed on an algebraic plane curve as a variety and applying Bernshtein's Theorem we are able to count the number of such squares. An algebraic plane curve defined by a polynomial of degree m inscribes either an infinite amount of squares, or at most (m4 - 5m2 + 4m)= 4 squares. Computations using computer algebra software lend evidence to the claim that this upper bound is sharp for generic curves. Earlier work on Toeplitz's conjecture has shown that generically an odd number of squares is inscribed on a smooth enough Jordan curve. Examples of real cubics and quartics suggest that there is a similar parity condition on the number of squares inscribed on some topological types of algebraic plane curves that are not Jordan curves. Thus we are led to conjecture that algebraic plane curves homeomorphic to the real line inscribe an even number of squares.
  • Suomen kansallisten päästövähennystoimien riskien ja kustannustehokkuuden arviointi
    (2011) Hast, Aira
     School of Science | Master's thesis
    According to the EU climate and energy package Finland should reduce greenhouse gas (GHG) emissions in national, non-trading sectors (non-ETS) at least 16 % below 2005 levels by 2020. In order to meet this target Finland has to implement GHG abatement activities. A situation where mitigation costs should be as low as possible is studied in this Thesis. To minimize the costs, mitigation activities with the best cost-efficiency should be chosen when forming optimal abatement portfolios. However, the amount of GHG reductions and costs are uncertain with every abatement activity and therefore portfolios involve risks to reduce emissions less than predicted or cause higher costs than estimated beforehand. The objective of this Thesis is to build portfolios which fulfil the reduction target. Each portfolio consists of activities that are chosen to be implemented in the examined timeline 2010-2020 and also the year they will be implemented. In this Thesis abatement activities are chosen among 17 independent mitigation actions. To form an optimal portfolio for different levels of GHG reductions a stochastic optimization model is built. The amounts of the risks related to the costs and reductions in different portfolios are then compared. The results from this analysis show that uncertainties in costs and reductions are almost equal in every examined efficient portfolio, when the reductions are gained by national mitigation actions. It also seems that the risk to reduce emissions less than expected cannot be lowered even if the expected value of costs is raised. The probability to meet the reduction target seems to depend strongly on the expected value of costs, so that a higher probability to meet the target involves higher costs. In addition to this it can also be seen that increasing probability to meet the target in higher standard of probability requires higher relative costs than in lower standard of probability because marginal abatement costs are increasing in function of gained reductions. The results show that some abatement actions are chosen in nearly all efficient portfolios while other actions are chosen extremely seldom. To meet the reduction target the Member States of EU can trade their non-ETS allocations. Two cases are compared to each other in this Thesis. In the first case the Member States have to meet the target by national mitigation actions, and in the second case the Member States can also trade non-ETS allocations in 2020. The possibility of trading allocations changes the set of implemented actions and postpones their optimal timings because optimization is done by minimizing the present value of overall costs. The portfolios that minimize overall costs are also studied so that comparison is made by examining how the costs, reductions gained by national mitigation actions and uncertainties related to them differ from each other in these two cases. The results prove that overall costs are approximately 10 % lower when the Member States can trade allocations. On the other hand, when trading allocations is possible, the risk to exceed the expected costs increases because the price of allocation unit is very uncertain. Sensitivity analysis is performed for cost minimizing portfolios in different cases so that overall costs and gained reductions are studied separately. Uncertainty in gained reductions is caused by the same sources in both cases. Yet, the variables causing uncertainty in costs are somewhat different in different cases.
  • Automatic Assessment in University-level Mathematics
    (2009) Ruokokoski, Jarno
     Perustieteiden korkeakoulu | Master's thesis
    In this thesis we study automatic assessment of randomized, university mathematics exercises. We use one system meant for this purpose and implement exercise collections from different mathematical topics. The selected topics of mathematics are calculus and graph theory. More of something about mathematical proving is presented. The topics were selected to fit for teaching university level mathematics in the department of automation and systems technology and the department of computer science and engineering. Calculus is assumed to be familiar, but graph theory are presented that understanding is possible without earlier knowledge. We wrote about 80 exercises for this thesis. We used the latest technical vations in automatic assessment in the writing of the exercises, and we tested received results in two courses, in which participate about 300 students during the semester 2008-2009. We will present the most interesting exercises, results of the experimentations and received feedback in this thesis.
  • Diskreetin maksimaalifunktion perusominaisuudet
    (2011) Ropponen, Jonatan
     School of Science | Master's thesis
    The Hardy-Littlewood maximal function is one of the most essential operators in real analysis and harmonic analysis but is not always sufficiently regular. The Hardy-Littlewood maximal function maps the functions of Lp-space into Lp-space but does not retain the continuity or Hölder continuity of functions, for instance. It would be beneficial to define a maximal function that is of the same magnitude as the Hardy-Littlewood maximal function and has the desired regularity properties. For this purpose, in this master's thesis we define the discrete maximal function, which also retains continuity and Hölder continuity. The discrete maximal function can be used in relatively general metric spaces. For the purposes of this work, the most important assumption of the examined metric space is that the used measure is doubling. Then the measures of spheres with the same center are proportional to each other. The measure being doubling enables the use of numerous tools for the analysis of metric spaces but is not too demanding as an assumption, since there are plenty of suitable spaces. The discrete convolution is used in defining the discrete maximal function. Similarly, the Hardy-Littlewood maximal function can be defined in real spaces with the ordinary convolution, but the ordinary convolution is not defined in general metric spaces. Of the properties of the discrete maximal function, we examine in particular Lp-properties and Hölder continuity. The determined properties are compared to the properties of the Hardy-Littlewood maximal function. We also show that the discrete maximal function is of the same magnitude as the Hardy- Littlewood maximal function. The methods used in the work include in particular results derived from the doubling measure and essential covering theorems, such as the Vitali and Whitney covering theorems. In this work, we examine the discrete maximal function in both the global and local case. We then notice that many of the properties are relatively similar in both cases. Therefore, the discrete maximal function can be limited to a subset of the examined space without losing important properties.
  • Bayesian Games for Analysis of Asymmetric Warfare Tactics
    (2011) Eskola, Olli
     School of Science | Master's thesis
    Roadside bombs have caused more than half of the casualties of the U.S. troops in Afghan war. The aim of this work is to examine the threats of the modern asymmetric warfare and to develop operations research based methods for advanced threat preparation. The main emphasis is on the roadside bombs. In this work, we are investigating the conflict between two forces. We define an influence diagram that contains the events relevant after the initialization of the bomb. The social support of the local people is included in the model because it is a key factor in asymmetric warfare. Following Bayesian game theory, both forces have two types and the types define the tactics of the forces. The types are given either as a joint common probability distribution or as subjective conditional probabilities. For the conditional probabilities, a theorem is formulated to define when they are mutually consistent. When the conditional probabilities are consistent, the common joint distribution can be calculated using the Bayes rule. When the subjective probabilities are mutually inconsistent, the Kullback-Leibner divergence measure is used for calculating the common joint probability distribution that is closest in terms of this measure, given the subjective probabilities. An algorithm solving the Bayesian Nash Equilibrium of the normal form game is programmed. The equilibrium shows the best response, i.e. the tactics adopted by players maximizing expected utility. By assumption, the utilities sum up to constant. The influence diagram combined with carefully elicited expert judgments provides a systematic approach for modeling the existing uncertainties related to the outcomes of the applied tactics. Several computational cases are examined. Based on these, the social support, for example, is more significant than the type probabilities of the forces.
  • BV-funktiot metrisissä mitta-avaruuksissa
    (2013) Lehtelä, Pekka
     Perustieteiden korkeakoulu | Master's thesis
    Functions of bounded variation, abbreviated as BV functions, define an important extension of the Sobolev functions, which is used in the calculus of variations and geometric measure theory. In the Euclidean case, BV functions are locally integrable functions, whose weak first partial derivatives are Radon measures. In the metric case BV functions are defined using relaxation as the closure of Lipschitz functions with respect to suitable convergence. In a similar manner the variation measure of a function is defined. An important part of the BV theory is the sets of finite perimeter, i.e. the sets whose characteristic functions are in BV. Thus the variation measure of the characteristic function defines the perimeter measure of the set. Main tools for the study of sets of finite perimeter are the isoperimetric inequality, which relates the measure of a set and its perimeter, and the coarea formula, which relates the variation measure of a function and the perimeters of its level sets. The perimeter of a set gives, in a sense, a measure for the boundary of the set. In the Euclidean case the perimeter of a set is given by n - 1 dimensional Hausdorff measure of the measure theoretic boundary of the set. As a main result of the thesis, a corresponding structure theorem is given in the metric setting. The theorem states that the perimeter of a set of finite perimeter is given by an integral of a bounded Borel function over the measure theoretic boundary of the set with respect to a Hausdorff type measure of codimension 1.
  • Cheap talk and cooperation in Stackelberg games
    (2012) Leppänen, Ilkka Johannes
     School of Science | Master's thesis
  • Classifying Individual-Level Migration Behavior Using Multivariate Methods
    (2010) Mangs, Karl Johan
     School of Science | Master's thesis
    The decision to migrate tends to be coupled with significant changes in the migrant's personal life. Consequently, the ability to forecast which individuals are about to migrate would offer companies an excellent opportunity to benefit from this fact in their prospecting and marketing. Targeting imminent migrants with well-timed and suitable offers would yield higher readership levels due to the more interested prospects. The aim of this thesis was to develop a method for classifying the adult Finnish population into one of the following categories: those who will migrate within the following six months, and those who will migrate at a later date. Utilizing the k-nearest neighbour algorithm, generalized linear models, support vector machines, optimally-pruned extreme learning machines, and their ensembles this thesis sought to analyze which approach offers the best classification result and whether there are significant differences regarding the practical implementations of these methods. Provided access to the full extent of the Population Information System database, this thesis was offered a premium opportunity to model the Finnish migration phenomenon through a large sample size and an extensive set of features. On the other hand, this created a demand for feature selection, which in this thesis was conducted based on a literature review. Moreover, the Delta Test was implemented as a method for pruning the selected shortlist from additional redundant features. The results revealed that there exists a subset of features that have a substantial impact on the classification accuracy. In addition, it was concluded that all of the implemented methods were capable of producing satisfactory results and that their performances were rather similar. However, considering that this thesis was able to highlight some very distinct practical differences regarding the methods, these findings provide valuable insights concerning future implementations.
  • Tiepäällysteiden ylläpitostrategioiden vertailu simulointimallilla
    (2011) Junes, Janne
     School of Science | Master's thesis
    Increasing traffic volume and decreasing maintenance budget force road agencies to develop new strategies for the road maintenance. This thesis studies road pavement maintenance strategies by mathematical modeling. Modeling focuses on high traffic volume main roads where pavement rutting is the main reason for resurfacing. Rutting is an inhomogeneous phenomenon, so that rut depth may vary considerably within a road section even if the age of the pavement and traffic volume were the same. This thesis compares road pavement maintenance strategies on high traffic volume roads. One strategy is patching which means that worst rutted spots are patched to wait for the actual resurfacing. Another strategy is to increase maximum rut depth allowed on the roads. The aim is to decrease the maintenance cost by delaying the actual resurfacing. This thesis develops a new approach to the research of road pavement maintenance strategies. A simulation model is built to model rutting and pavement maintenance. Strategies are compared using the model without field tests on roads. A rutting model to produce realistic inhomogeneous rut depth profiles is developed for the simulation model. The strategies are converted to decision rules related to rut depth. The simulation model executes maintenance to a virtual road based on the decision rules. The model features a new condition-based criterion to stop the simulation run. This new criterion helps to provide more comparable information about the annual maintenance costs of different strategies than using a fixed number of simulation years. Strategies are compared based on the results from the simulation model which gives an extensive amount of information about the effects of different strategies. New methods and tools as well as new information about the effects of different strategies to the maintenance costs and pavement condition are obtained as the result of this thesis. Increasing the average rut depth of resurfacing sites by one millimeter generates average savings of about 10 percent in annual maintenance cost. About 15 percentage savings are reached by patching. Combination of patching and increased average rut depth generates about 20 percent savings. A disadvantage is that the service level experienced by road user decreases due to patches and increased rut depth. The results of this thesis can be used by the Finnish Transport Agency to update the maintenance policy and by road districts to support maintenance programming.
  • Computing Nash equilibrium in a telecommunication market model with different cognitive radio scenarios
    (2013) Niemenmaa, Eerno
     Perustieteiden korkeakoulu | Master's thesis
    In this thesis we are investigating how the different cognitive radio scenarios affect the operators, the customers and the government. We have constructed a telecommunication market model to find out the consequences. We have constructed a two-stage game based on this model. In the first stage operators buy capacity. That means licensing spectrum and maintaining the network technology. In the second stage the operators set prices to their service. The special properties in our game are limited availability of capacity and uncertainty of demand. The limited availability of capacity is simulated by using a quadratic cost function. The uncertainty of demand is generated by adding a random variable to the demand function. The main theme of this thesis is comparing different scenarios to each other. For each scenario we developed functions to represent the additional capacity. Finally we compare the results by using indexes created to represent the utility of different sides. Solving the game is started by using analytical methods with general number of players. However, the complexity of dependencies prevents solving game analytically. We continue with numerical methods. With numerical methods we primarily are searching pure strategy Nash equilibrium. It turns out that pure strategy equilibrium does not exist generally. To solve mixed strategy equilibrium, we form a normal form game to region that our algorithm keeps oscillating and solve its mixed strategy equilibrium. According to the results of this thesis, one of the suggested changes in rules would be equally profitable for operators, but remarkably better by the government's and the customers' point of view.
  • Constructing a model for Search and Rescue in Gulf of Finland
    (2012) Torabihaghighi, Fatemeh
     School of Science | Master's thesis
    The Finnish waters of Gulf of Finland provide a perfect setting for recreational boating activities. Each year, a considerable number of boaters are in need of medical or technical assistance. An extensive sea rescue system is operated by the Finnish Boarder Guard and the Finnish Lifeboat institution, to perform Search and Rescue (SAR) and assistance operations to safeguard life at sea. The main questions are related to the required number, type and location of the rescue vessels to attain a desired service level. This study evaluates the performance of Search and Rescue for recreational vessel in Gulf of Finland by making a simulation model of the sea rescue services. The discrete-event simulation and the Monte Carlo technique are applied to make the simulation model. The model is based on an extensive incident data analysis of historic incidents in the Finnish Gulf of Finland. The current version of the model still needs significant improvements before it can provide trustworthy decision support, but the preliminary results indicate that the approach works in principle and that valuable information about the system performance can be obtained from such an approach.
  • Cross-Border Exits and Private Equity Fund Performance
    (2012) Tapola, Niklas
     School of Science | Master's thesis
    European venture capital and private equity industry has grown rapidly and experienced a dramatic internationalization during the past decades. The purpose of this study is to study cross-border exits, which include investors selling a portfolio company to a foreign firm or listing the company into a foreign stock exchange, and to examine what role these exits play in private equity and venture capital fund performance. Earlier research has identified many potential determinants of venture capital fund performance. The older and more experienced venture capitalists are commonly able to yield higher returns for the investors than younger investors, similarly large funds typically outperform small ones, and funds with precise industry or geographic focus normally outperform the broader funds. The earlier research has also pointed out several motives to pursue international exits. With trade sales, ventures may find strategic buyers outside their home country. With initial public offerings, securities are commonly worth more in countries with more developed capital markets and better legal protection of investors. Venture capitalists are needed to certify the quality of ventures especially with cross-border exits. Based on previous research, the study develops hypotheses that funds active in cross-border yield higher returns, and the positive effects of cross-border exits are more pronounced when the venture originates from a country with less developed local exit conditions. The hypotheses are tested employing a novel dataset covering European venture capital funds incepted between years 1982 and 2004. The potential sample biases in performance reporting are controlled using Heckman correction. The results find support that venture capital funds active in cross-border exits yield higher returns to the investors than the funds less active in cross-border exits. However, the results do not find support that the local exit conditions would affect cross-border exit performance significantly.
  • Customer Churn Modeling Using Sales System Data
    (2014) Ylilammi, Kimi
     Perustieteiden korkeakoulu | Master's thesis
    Acquiring new customers is more expensive and challenging than retaining existing customers. Therefore, companies should take care of their existing customers in order to retain them. These so-called retaining actions should only be used for unsatis ed customers who are about to transfer their business somewhere else. In this thesis we analyze the Finnish National Opera's (FNO) customers and develop a model to identify churning customers. Furthermore, additional continuous variables are predicted from the customers, e.g. the number of tickets purchased next year. These additional predictions are intended to further increase the efficiency of retaining actions by being able to personalize and more accurately target individual customer needs. A random forest model is developed and benchmarked to other popular models in churn modeling. All the models are built using the FNO's sales system data. Random forests outperform other models in churn modeling and in most of the continuous variable prediction problems. The random forest model has high accuracy when used for churn modeling, but its performance is poorer when predicting continuous variables. Our model is the most accurate when predicting how many days in advance customers purchase their tickets. The churn model is able to predict the churn of all customers with equal accuracy, and while it is accurate in identifying churners, it is not able to find all of them
  • Development and Communication of Short Term Sales Forecasts
    (2013) Nysten, Tony
     School of Science | Master's thesis
  • Divergence-free Interpolation Using Polynomial Splines
    (2012) Perkkiö, Lauri
     School of Science | Master's thesis
    The work is motivated by particle trajectory tracking in a magnetic field, which is always divergence-free. Given a discrete set of vector data points (which may represent a magnetic field), we wish to construct a divergence-free interpolant. Some well-known divergence-free methods already exist, but we aim at one which is smooth and computationally fast. The interpolant is constructed by fitting a piecewise polynomial into the data using a vector potential as an intermediary step. Unwanted oscillation, which the high-order polynomials are prone to, is removed by minimizing the integral of second derivatives of the interpolant. As a result, we need to solve a global problem first, but afterwards the evaluation is straightforward and local. The method is tested in two dimensions and regular grids, but in three dimensions and/or irregular grids some extra optimization is needed to make the method applicable.
  • Dyadiset kuutiot homogeenisen tyypin avaruudessa
    (2012) Korvenpää, Janne
     School of Science | Master's thesis
    In this thesis, we generalize the system of Euclidean dyadic cubes used in harmonic analysis to a space of homogeneous type, i.e. a quasi-metric space with a doubling measure. The essential properties of the dyadic cubes are that they form a tree structure such that any two of them are either disjoint or one is contained in the other, and that each generation of cubes covers the whole space excluding a possible set of measure zero. In addition, dyadic cubes are not too far away from balls in the sense that they are bounded by balls of the same magnitude from inside and outside. The most central applications of dyadic cubes are the dyadic Calderón-Zygmund decomposition and the dyadic maximal function. They are used in harmonic analysis and they do not significantly differ from their Euclidean counterparts.
  • Ensihoitopalvelujen simulointi- ja optimointimalli
    (2011) Luukkonen, Teppo
     School of Science | Master's thesis
    The daily workload of an ambulance crew can be significant if there are several medical emergencies during the day. However, the workload can also be low if there are few emergency calls. Thus the workload for single paramedic can vary greatly. Excessive overtime work exhausts the paramedics but on the other hand paramedics' unused work hours are wasted resources. Thus balancing of paramedics' workloads is needed. This is possible by simulating workloads of an ambulance and organizing the work shifts according to the simulated workloads. Three simulation models are developed to estimate the daily workloads of an ambulance. One ambulance unit is considered and the purpose of the simulation is to construct work days similar to the data from the ambulance in question. Non-parametric simulation is based on the tasks previously carried out by the ambulance. Parametric simulation applies theoretical distributions in the given data. A queuing model simulates the tasks given to the ambulance by an emergency centre. Work schedules are planned according to the simulated workloads. The aim is to distribute the workload evenly among all paramedics. Thus mixed integer linear problem is formulated in accordance with the constraints required by the paramedics and the collective labour agreement. Work shifts are generated by the optimization model using simulated workloads. Then the solutions are compared to each other by utilising stochastic dominance and several simulations. Alternative good solutions are compared by examining expected value, variance, maximum and minimum values, and VaR and CVaR risk measures of the objective function. The results indicate that the simulation and optimization models generate excellent alternative work shift schedules to the given problem. However, the simulation models and other methods used in the process require further research so that the reliability of the given solutions is improved.
  • Engine exhaust plume interactions with a planetary surface
    (2014) Kahila, Heikki
     Perustieteiden korkeakoulu | Master's thesis