Master of Science / Department of Agricultural Economics / Glynn Tonsor / This decade has seen movements in commodity futures markets never seen before. There are many factors that have intensified price movements and volatility behavior. Those factors likely altering supply and demand include governmental policy within and outside of the U.S, weather shocks, geopolitical conflicts, food safety concerns etc. Whatever the reasons are for price movements it is clear that the volatility behavior in commodity markets constantly change, and risk managers need to use current and efficient tools to mitigate price risk. This study identified market structural breaks of realized volatility in corn, wheat, soybeans, live cattle, feeder cattle and lean hogs futures markets. Furthermore, this study analyzes the forecasting performance of implied volatility, historical volatility, a composite approach and a naïve approach as forecasters of realized volatility. The forecasting performance of these methods was analyzed in the full period of time of our weekly data from January 1995 to April 2014 and in each identified market regime for each commodity. Previous research has analyzed forecasting performance of implied volatility, a time series alternative and a composite method. However, to the best of my knowledge, they have not worried about market structural breaks in the data that might influence the performance of the mentioned forecasting methods in different periods of time. Overall, results indicate that indeed there are multiple market structural breaks present in the volatility datasets across all six commodities. We found differences in the forecasting performance of the analyzed methods when individual market regimes were analyzed. There seems to be evidence that corroborates the idea in the literature about the superiority of implied volatility over a historical volatility, a composite approach and a naïve approach. Additionally, implied volatility encompassed all the information contained in the historical volatility and the naïve measure across each identified market regime in all six commodities. Our results show that when both implied volatility and historical volatility are available, the benefit of combining those measures into a composite forecasting approach is very limited. Our results hold true for a short term 1 week ahead realized volatility forecast. It would be of interest to see how results vary for longer forecasting time horizons.

The problem of profiled school scheduling is important for Lithuanian schools and for similar schools in many other countries. No polynomial time methods are known for this problem. The objective of this PhD thesis is to investigate heuristic methods for optimization of profiled school schedules. The convenience of application in real-life situations is provided by the vector optimization approach using platform-independent software implementation. The task of experimental investigation is to select such parameters of heuristic methods that minimize expected deviation from the optimum. Four optimization methods were regarded: Local Deterministic (LD), Local Randomized (LR), Simulated Annealing (SA), and SA with parameters optimized using the Bayesian approach (BA). The composite method of AM and BA provided the best results. In Chapter 1 of the dissertation, various aspects of work of optimization methods as well as popular program languages suitable for school schedule optimization are analyzed. Literature about school scheduling is analysed. In Chapter 2, conclusions are drawn how the optimization of heuristic parameters influences the speed and accuracy of finding the optimal solution. A technical rating analysis of popular schedule programs is made and technical disadvantages are listed. Criteria for evaluating the quality of results are proposed that include heuristic parameters in search of optimal schedules. Recommendations are states how to assess the choice and... [to full text] / Profiliuotos mokyklos tvarkaraščio kūrimas yra aktualus uždavinys tiek Lietuvoje, tiek kitose šalyse. Nėra žinoma polinominių būdų šiai problemai spręsti. Pagrindinis šios daktaro disertacijos objektas yra ištirti euristinius metodus, skirtus profiliuotos mokyklos tvarkaraščio optimizavimui. Tvarkaraščio formavimo kriterijai, kurie yra reikalingi realiame gyvenime, nustatomi vektorinio optimizavimo metodais bei realizuojami nuo operacinės sistemos nepriklausoma programine įranga. Eksperimentinių tyrimų uždavinys – surasti tokius euristinių metodų parametrus, kurie minimizuotų numatytą nuokrypį. Disertacijoje aprašomi keturi optimizavimo metodai: lokalus determinuotas (LD), lokalus atsitiktinis (LA), atkaitinimo modeliavimo (AM) ir AM parametrų optimizavimas naudojant Bayes (BA) metodą. Kombinuotas AM ir Bayes metodas duoda geriausius rezultatus. Pirmajame apžvelgiama su mokyklos tvarkaraščių formavimu susijusi literatūra. Analizuojami tinkamiausi optimizavimo metodų darbo aspektai. Analizuojamos populiariausios programavimo kalbos, tinkančios kurti mokyklų tvarkaraščių optimizavimo programą. Antrajame skyriuje formuluojamas profiliuotų mokyklų tvarkaraščio kūrimo matematinis modelis. Analizuojami profiliuotose mokyklose naudojami euristiniai parametrai. Atlikti populiarių tvarkaraščių programų vertinimai ir analizės. Įvardijami šių programų trūkumai. Trečiajame skyriuje, remiantis 2 skyriaus analize ir išvadomis, pateiktas profiliuotos mokyklos lanksčios tvarkaraščio... [toliau žr. visą tekstą]

Profiliuotos mokyklos tvarkaraščio kūrimas yra aktualus uždavinys tiek Lietuvoje, tiek kitose šalyse. Nėra žinoma polinominių būdų šiai problemai spręsti. Pagrindinis šios daktaro disertacijos objektas yra ištirti euristinius metodus, skirtus profiliuotos mokyklos tvarkaraščio optimizavimui. Tvarkaraščio formavimo kriterijai, kurie yra reikalingi realiame gyvenime, nustatomi vektorinio optimizavimo metodais bei realizuojami nuo operacinės sistemos nepriklausoma programine įranga. Eksperimentinių tyrimų uždavinys – surasti tokius euristinių metodų parametrus, kurie minimizuotų numatytą nuokrypį. Disertacijoje aprašomi keturi optimizavimo metodai: lokalus determinuotas (LD), lokalus atsitiktinis (LA), atkaitinimo modeliavimo (AM) ir AM parametrų optimizavimas naudojant Bayes (BA) metodą. Kombinuotas AM ir Bayes metodas duoda geriausius rezultatus. Pirmajame apžvelgiama su mokyklos tvarkaraščių formavimu susijusi literatūra. Analizuojami tinkamiausi optimizavimo metodų darbo aspektai. Analizuojamos populiariausios programavimo kalbos, tinkančios kurti mokyklų tvarkaraščių optimizavimo programą. Antrajame skyriuje formuluojamas profiliuotų mokyklų tvarkaraščio kūrimo matematinis modelis. Analizuojami profiliuotose mokyklose naudojami euristiniai parametrai. Atlikti populiarių tvarkaraščių programų vertinimai ir analizės. Įvardijami šių programų trūkumai. Trečiajame skyriuje, remiantis 2 skyriaus analize ir išvadomis, pateiktas profiliuotos mokyklos lanksčios tvarkaraščio... [toliau žr. visą tekstą] / The problem of profiled school scheduling is important for Lithuanian schools and for similar schools in many other countries. No polynomial time methods are known for this problem. The objective of this PhD thesis is to investigate heuristic methods for optimization of profiled school schedules. The convenience of application in real-life situations is provided by the vector optimization approach using platform-independent software implementation. The task of experimental investigation is to select such parameters of heuristic methods that minimize expected deviation from the optimum. Four optimization methods were regarded: Local Deterministic (LD), Local Randomized (LR), Simulated Annealing (SA), and SA with parameters optimized using the Bayesian approach (BA). The composite method of AM and BA provided the best results. In Chapter 1 of the dissertation, various aspects of work of optimization methods as well as popular program languages suitable for school schedule optimization are analyzed. Literature about school scheduling is analysed. In Chapter 2, conclusions are drawn how the optimization of heuristic parameters influences the speed and accuracy of finding the optimal solution. A technical rating analysis of popular schedule programs is made and technical disadvantages are listed. Criteria for evaluating the quality of results are proposed that include heuristic parameters in search of optimal schedules. Recommendations are states how to assess the choice and... [to full text]