Πλοήγηση ανά Επιβλέπων "Zazanis, Michael"
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω
Τώρα δείχνει 1 - 4 από 4
- Αποτελέσματα ανά σελίδα
- Επιλογές ταξινόμησης
Τεκμήριο Probabilistic models in Risk TheoryAlexiou, Panagiotis; Αλεξίου, Παναγιώτης; Athens University of Economics and Business, Department of Statistics; Zazanis, MichaelIt is common practice for insurance companies to give dividends to their shareholders. Many papers have been written on dividends policies. It has been found that under some reasonable assumptions the optimal policy is to follow the so called constant barrier dividends policy also known in Risk theory as de Finetti model. However soon become apparent that this model is not "perfect" as questions and problems emerge from its application and that some "kind" of modications are necessary. In this spirit, this thesis extends the de Finetti model in order to include cases with barriers dividends policies which are modeled by diffusions. The approach is axiomatic and was motivated by the classical de Finetti model. We show that the de Finetti models with general (diffusion) barriers are well posed that is they exist and are unique, or in other words that there exist unique stochastic processes that evolve according to our conditions. When we say unique stochastic processes we mean up to the degree of indistinguishability. We consider de Finetti models with one general barrier meaning that when the reserves of the insurance company reach a "particular" level, which also depends upon a diffusion process, then the company goes bankrupt. We also consider de Finetti models with two general barriers, that is when the reserves of the insurance company reach the level of the lower barrier, which also depends on a di¤usion process, then the insurance company has the option to borrow money and continue it s function. We derive differential equations with appropriate boundary conditions, the solution of which gives the quantities for which we are interesting. More specifically we find differential equations with appropriate boundary conditions, the solution of which gives the moments of the discounted dividends, the moments of the discounted financing, the Laplace transform of the time of ruin, the Laplace transform of the joint distribution of the time of ruin and the discounted dividends and the Laplace transform of the joint distribution of the discounted dividends and the discounted financing. We apply the formulas in special cases and more specifically in cases where the reserves process follows a Brownian motion, a Geometric Brownian motion and an Ornstein Uhlenbeck process (also see Gerber, H.U. and Shiu, E.S.W.([71],[72])). Next we work on another important issue, which is the situation of insurance companies cooperation. We consider this issue from the perspective of a particular insurance company. We are interesting to look at parameters which are vital to the decisions of the company. Among these parameters very important role we consider to play the probability of survival in a particular cooperation and the shares that will be given to the shareholders during this cooperation. We find differential equations with appropriate boundary conditions the solution of which will give:• The moments of the discounted dividends and the discounted financing.• The Laplace transform of the joint distribution of the time of ruin and the discounted• dividends.• The Laplace transform of the discounted dividends.• The Laplace transform of the time of ruin.• The Survival probability for one of the two insurers.We apply these results in two models:(I) The Lundberg - de Finetti model.(II) The de Finetti - de Finetti model.We show how an insurance company can use the above results for policy making purposes. We also mention possible ways to extend the above considerations to various other models.Τεκμήριο Queues, quality control and success runs: methodology and applications in productionMytalas, George; Μύταλας, Γεώργιος; Athens University of Economics and Business, Department of Statistics; Giannakopoulos, Thanasis; Fragkos, N.; Zazanis, MichaelThis thesis deals with stochastic models in production systems and the research conducted followed three directions: queueing models in production, quality control, and run statistics. These are presented independently of each other but there are clear interconnections between them. The first part of the thesis (Chapters 1 and 2) deals with success runs in Markov--dependent binary trials. There we develop a new methodology based on a Central Limit Theorem for additive functionals of Markov chains in order to establish the asymptotic normality of the joint distribution of different types of success runs. The approach is based on expressing the number of runs in terms of the number of visits of an appropriately defined Markov chain to certain states, or in terms of the number of state transitions of the chain. Then we take advantage of the connection of the asymptotic covariance matrix for such functionals to the recurrent potential matrix (fundamental matrix) of this chain. The last step is to express the elements of the potential matrix in terms of the mean transition times between states for the Markov chain and thus to obtain explicit, closed form expressions for the limiting covariance matrix. In many cases this has led to new results. In the second part (Chapter 3) we analyze the performance of several continuous sampling plans such as CSP-1, CSP-2, CSP-3, CSP-C, and Tightened Multi-Level (3) under the assumption that the quality of the examined items exhibits markovian dependence while probability sampling is used during the sampling inspection phase for each sampling plan. The assumption of markovian dependence in the quality process is essential since the Bernoulli assumption which is often used for analytical convenience fails to capture the essential features of these plans that aim to capture shifts in the quality process. We present a method for computing the exact probability generating function for the joint number of items (defective and nondefective) both for the continuous inspection phase and for the sampling inspection phase, using a markovian analysis. For each plan we derive exact results for the Average Fraction Inspected (AFI), the Fraction Defective Outgoing (FDU), the Average Outgoing Quality (AOQ), both for replacement and for non-replacement policies and finally we present numerical results that provide comparisons between the plans. The third part (Chapters 4, 5, and 6) considers queueing problem arising in the analysis of manufacturing systems. In chapter 4, we examine the operation of a quality control unit that uses a Continuous Sampling Plan (CSP) policy, placing particular emphasis on the delay aspects of the inspection station and on the effects of the inspection process on flow times. In particular, we examine markovian models of the inspection station and analyze the delay and queueing aspects of the CSP policy assuming exponential inspection times using matrix geometric techniques. We also examine the indirect effects on the inspection process on flow times due to the modulation of the downstream arrival process by the inspection process and the corresponding increase of variability. In chapter 5 a queueing model with Poisson arrivals, random processing batches, and vacations is examined, specifically based on the Queue-Cart model of Coffman and Gilbert [37] and we extend their results to the case where the cart's capacity varies stochastically and customers arrive in batches. We also examine processing batch distributions with unbounded support and provide a solution via Wiener-Hopf techniques. In all cases care is taken in the analysis in order to obtain the steady state distribution without the assumption that the service and cart delivery time distributions are light-tailed. In fact, our results are obtained under the natural conditions of finite first moments, together with the stability condition which guarantees the existence of a stationary version of the process. Finally, in chapter 6, we examine an M/G/1 queueing model with batch arrivals subject to disasters. The model includes a randomized vacation policy and repair times after disasters. By applying the supplementary variables method various performance criteria were obtained such as the queue length distribution, the Laplace transform of the waiting time, the Laplace transform of the busy period, and the probability that a typical customer suffers a disaster.Τεκμήριο Ruin theory problems in simple SDE models with large deviation asymptoticsΜπουγιουκλή, Ευσταθία; Bougioukli, Efstathia; Athens University of Economics and Business, Department of Statistics; Frangos, Nikolaos; Chadzikonstantinidis, Efstathios; Yannacopoulos, Athanasios; Vakeroudis, Stavros; Glasserman, Paul; Schmidli, Hanspeter; Zazanis, MichaelΕξετάζουμε προβλήματα πιθανοτήτων σχετικά με τη συμπεριφορά απλών γραμμικών στοχαστικών διαφορικών εξισώσεων με εκθετικά όρια, που σχετίζονται με προβλήματα που προκύπτουν στη θεωρία κινδύνου και στα μοντέλα asset και liability στα pension funds. Το πρώτο μοντέλο που εξετάζουμε, στο Κεφάλαιο 2, είναι μια διαδικασία Ornstein-Uhlenbeck (OU) που περιγράφεται από τη Στοχαστική Διαφορική Εξίσωση dX_t=μ X_t dt+σdW_t με 〖 X〗_0=x_0 δεδομένο, όπου μ>0 και {W_t } είναι standard κίνηση Brown. Αυτό το μοντέλο προκύπτει ως μια προσέγγιση διάχυσης των μοντέλων θεωρίας κινδύνου στα οποία τα «ελεύθερα» αποθεματικά τοκίζονται. Το ερώτημα που τίθεται είναι αυτό του προσδιορισμού της πιθανότητας η διαδικασία να «χτυπήσει/συναντήσει» μια κάτω ντετερμινιστική καμπύλη υ_0 e^βt ή/και μια άνω καμπύλη u_0 e^αt, υποθέτοντας αρχικά ότι τα «ελεύθερα» αποθέματα βρίσκονται μεταξύ αυτών των τιμών, δηλ. 0< υ_0< x_0< u_0 και ότι β<μ<α. Εξετάζονται τόσο το «πρόβλημα πιθανοτήτων καταστροφής» του πεπερασμένου ορίζοντα για τον προσδιορισμό της πιθανότητας πρόσκρουσης στο όριο εντός ενός πεπερασμένου ορίζοντα, όσο και η πιθανότητα άπειρου ορίζοντα. Αυτό το πρόβλημα μπορεί φυσικά να διατυπωθεί με όρους PDE δεύτερης τάξης με καμπύλα (εκθετικά) όρια στο επίπεδο και να λυθεί αριθμητικά. (Μια εναλλακτική προσέγγιση που περιλαμβάνει ένα επιχείρημα αλλαγής χρόνου (time change argument)αναφέρεται εν συντομία στο Κεφάλαιο 2.) Το κύριο μέρος της ανάλυσης περιλαμβάνει τεχνικές Large Deviations και ειδικότερα την προσέγγιση των Wentzell-Freidlin προκειμένου να ληφθούν λογαριθμικές ασυμπτωτικές για την πιθανότητα «πρόσκρουσης» είτε του κάτω είτε του άνω ορίου. Οι ασυμπτωτικές χαμηλού θορύβου ισχύουν όταν η διακύμανση σ είναι μικρή και ως εκ τούτου το γεγονός να χτυπήσει οποιοδήποτε όριο είναι σπάνιο. Ο εκθετικός ρυθμός που χαρακτηρίζει την πιθανότητα προκύπτει με την επίλυση ενός variational προβλήματος το οποίο δίνει επίσης το «μονοπάτι προς την καταστροφή». Ξεκινάμε με μια προσεκτική και λεπτομερή ανάλυση του προβλήματος του πεπερασμένου ορίζοντα της πρόσκρουσης σε ένα κατώτερο όριο. Το πρόβλημα του άπειρου ορίζοντα τόσο για το «χτύπημα» του κάτω όσο και του άνω εκθετικού ορίου αντιμετωπίζεται χρησιμοποιώντας την προσέγγιση των transversality conditions του λογισμού μεταβολών. Επιπλέον, εξετάζεται η διαδικασία OU με γενικότερο συντελεστή γραμμικής μετατόπισης, δηλαδή η διαδικασία που προκύπτει από την SDE dX_t=(μ X_t+r)dt+σdW_t με το άνω εκθετικό όριο u_0 e^αt (με 0<μ<α).Θεωρούμε επίσης, στο τέλος του Κεφαλαίου 2, το πρόβλημα δύο ανεξάρτητων OU διαδικασιών που προκύπτουν από τα SDEs dX_t=α X_t dt+σdW_t, 〖dY〗_t=β Y_t dt+bdV_t, με δεδομένα X_0=x_0 , Y_0=y_0 . Επίσης, οι {W_t }και {V_t } είναι ανεξάρτητες κινήσεις Brown. Αν α>β και 〖 x〗_0> y_0 τότε, στην απουσία θορύβου, θα ισχύει ότι X_t> Y_t για όλα τα t>0. Εξετάζουμε, την πιθανότητα να συναντηθούν οι δύο διαδικασίες χρησιμοποιώντας ξανά την προσέγγιση των Wentzell-Freidlin. Οι βέλτιστες διαδρομές που ακολουθούνται από τις δύο διαδικασίες και ο χρόνος συνάντησης T καθορίζονται με την επίλυση ενός variational προβλήματος χρησιμοποιώντας transversality conditions. Είναι ενδιαφέρον ότι το ίδιο μοντέλο όταν θεωρήσουμε μια συσχέτιση μεταξύ των δύο κινήσεων Brown παρουσιάζει πιο περίπλοκη συμπεριφορά εάν ο συντελεστής συσχέτισης υπερβαίνει ένα ορισμένο όριο. Αυτή η τελευταία περίπτωση συζητείται στο κεφάλαιο 4.Στο Κεφάλαιο 3 εξετάζεται ένα αντίστοιχο πρόβλημα που περιλαμβάνει μια Γεωμετρική κίνηση Brown που περιγράφεται από τη SDE dX_t=μ X_t dt+σX_t dW_t με X_0=x_0 , μαζί με ένα άνω και ένα κάτω εκθετικό όριο. Και πάλι, χρησιμοποιείται η θεωρία των Wentzell-Freidlin. Σε αυτήν την περίπτωση, ωστόσο, μια ακριβής λύση είναι δυνατή, και επομένως μπορούμε να αποκτήσουμε μια ιδέα για την ακρίβεια των λογαριθμικών ασυμπτωτικών που προτείνουμε. Όπως αναμενόταν, όταν η διακύμανση σ γίνει μικρότερη, η εκτίμηση βελτιώνεται. Αναφέρεται επίσης η περίπτωση δύο συσχετιζόμενων γεωμετρικών κινήσεων Brown. Αυτά τα μοντέλα είναι εμπνευσμένα από το μοντέλο των Gerber και Shiu για τα assets & liabilities στα pension funds.Στο Κεφάλαιο 4, εκτός από την επανεξέταση του προβλήματος δύο διαδικασιών Ornstein-Ulhenbeck με την παρουσία συσχέτισης, εξετάζουμε επίσης εν συντομία διαδικασίες OU με χρονικά μεταβαλλόμενη διακύμανση, που προκύπτουν από την SDE dX_t=μ X_t dt+σ(t)dW_t . Το πρόβλημα «πρόσκρουσης» που εξετάζουμε έχει κάτω εκθετικό όριο και άπειρο ορίζοντα. Το πρόβλημα μεταβλητότητας που προκύπτει από τη μέθοδο Wentzell-Freidlin είναι επιλύσιμο. Ωστόσο, η εξίσωση που δίνει τον βέλτιστο χρόνο «πρόσκρουσης» μπορεί να μην έχει μοναδική λύση. Επιλύουμε ένα παράδειγμα αυτού του προβλήματος αριθμητικά για να επεξηγήσουμε την προσέγγιση.Τεκμήριο Synchronized and gated queueing models(30-11-2022) Πινότση, Δήμητρα; Pinotsi, Dimitra; Athens University of Economics and Business, Department of Statistics; Pavlopoulos, Charalampos; Kyriakidis, Epaminondas; Besbeas, Panagiotis; Livada, Alexandra; Zhao, Yiqiang Q.; He, Qi-Ming; Zazanis, MichaelΗ διατριβή πραγματεύεται ζητήματα σχετιζόμενα με θεωρία ουρών που αφορούν σε συστήματα με θύρες με άπειρο πλήθος εξυπηρετούντων καθώς και συστήματα με συγχρονισμένες αφίξεις. Τα συστήματα με θύρες με άπειρο πλήθος εξυπηρετητών έχουν εφαρμογές σε βιομηχανικές διαδικασίες καθώς και σε τηλεπικοινωνιακά συστήματα. Λόγω του μηχανισμού της θύρας δεν είναι θεμιτό να γίνει μια ακριβής ανάλυση και γι’ αυτό εφαρμόζονται αναλυτικές μέθοδοι (ή προσομοίωση). Το γεγονός του άπειρου πλήθους των εξυπηρετούντων και η παρουσία του μηχανισμού θύρας κάνει το ερώτημα της ευστάθειας του συστήματος ενδιαφέρον ζήτημα.Η ευστάθεια της ουράς M/G/∞ με πύλη, διερευνάται με τη χρήση ενός κριτηρίου ολίσθησης των Foster–Lyapunov βάσει του οποίου αποδεικνύεται ότι το να είναι πεπερασμένη η πρώτη ροπή της κατανομής των χρόνων εξυπηρέτησης είναι ικανή και αναγκαία συνθήκη για τη θετική επανάληψη του συστήματος.Ένα σύστημα που αποτελείται από m ανεξάρτητους, παράλληλους και εκθετικούς εξυπηρετούντες, με ανάλογη ντετερμινιστική εισροή διερευνήθηκε ως μια τροποποίηση του Flatto-Hahn-Wright μοντέλου της θεωρίας ουρών, το οποίο αντίθετα με το αρχικό εξελίσσεται σε πιο εύκολα διαχειρίσιμο αναφορικά με την ανάλυσή του. Εστιάζουμε στην ευσταθή κατανομή του χρόνου που χρειάζονται οι πελάτες, η οποία επιτυγχάνεται με τη χρήση μιας Μαρκοβιανής προσέγγισης εισαγωγής σε συνδυασμό τον τύπο αντιστροφής.Τέλος, πραγματοποιήθηκε μια λεπτομερής ανάλυση της περιόδου αιχμής ενός συστήματος M/G/∞. Αναλύθηκε η διάρκεια του σταδίου μιας Μαρκοβιανής αλυσίδας παρέχοντας πλήθος λύσεων για την περίπτωση που έχουμε μικρή κίνηση. Και στις δύο περιπτώσεις τα τελικά αποτελέσματα εξαρτώνται από τη λύση άπειρων γραμμικών συστημάτων