Raman spectroscopy has been successfully employed in the classification of breast pathologies involving basis spectra for chemical constituents of breast tissue and resulted in high sensitivity (94%) and specificity (96%) (Haka et al, 2005). Motivated by recent developments in nonparametric regression, in this work, we adapt stacking, boosting, and dynamic ensemble learning into a nonparametric regression framework with application to Raman spectroscopy analysis for breast cancer diagnosis. In Chapter 2, we apply compound estimation (Charnigo and Srinivasan, 2011) in Raman spectra analysis to classify normal, benign, and malignant breast tissue. We explore both the spectra profiles and their derivatives to differentiate different types of breast tissue. In Chapters 3-5 of this dissertation, we develop a novel paradigm for incorporating ensemble learning classification methodology into a nonparametric regression framework. Specifically, in Chapter 3 we set up modified stacking framework and combine different classifiers together to make better predictions in nonparametric regression settings. In Chapter 4 we develop a method by incorporating a modified AdaBoost algorithm in nonparametric regression settings to improve classification accuracy. In Chapter 5 we propose a dynamic ensemble integration based on multiple meta-learning strategies for nonparametric regression based classification. In Chapter 6, we revisit the Raman spectroscopy data in Chapter 2, and make improvements based on the developments of the methods from Chapter 3 to Chapter 4. Finally we summarize the major findings and contributions of this work as well as identify opportunities for future research and their public health implications.

The imminent emergence of the hydrogen fuel industry has resulted in an urgent mandate for very specific material testing. Although storage of pressurized hydrogen gas is both practical and attainable, demands for increasing storage pressures (currently around 70 MPa) continue to present unexpected material compatibility issues. It is imperative that materials commonly used in gaseous hydrogen service are properly tested for hydrogen embrittlement resistance. To assess material behavior in a pressurized hydrogen environment, procedures were designed to test materials for susceptibility to hydrogen embrittlement. Of particular interest to the field of high-pressure hydrogen in the automotive industry, austenitic stainless steels SUS 316 and 316L were used to validate the test programs. Tests were first performed in 25 MPa helium and hydrogen at room temperature and at -40°C. Tests in a 25 MPa hydrogen atmosphere caused embrittlement in SUS 316, but not in 316L. This indicated that alloys with higher stacking fault energies (316L) are more resistant to hydrogen embrittlement. Decreasing the test temperature caused slight embrittlement in 316L and significantly enhanced it in 316. Alternatively, a second set of specimens was immersed in 70 MPa hydrogen at 100°C until reaching a uniform concentration of absorbed hydrogen. Specimens were then loaded in tension to failure to determine if a bulk saturation of hydrogen provided a similar embrittling effect. Neither material succumbed to the effects of gaseous pre-charging, indicating that the embrittling mechanism requires a constant supply of hydrogen at the material surface rather than having bulk concentration of dissolved hydrogen. Permeation tests were also performed to ensure that hydrogen penetrated the samples and to develop material specific permeation constants. To pave the way for future work, prototype equipment was constructed allowing tensile or fatigue tests to be performed at much higher hydrogen pressures. To determine the effect of pressure on hydrogen embrittlement, additional tests can be performed in hydrogen pressures up to 85 MPa hydrogen. The equipment will also allow for cyclic loading of notched tensile or compact tension specimens for fatigue studies.

Credit scoring is very important process in banking industry during which each potential or current client is assigned credit score that in certain way expresses client's probability of default, i.e. failing to meet his or her obligations on time or in full amount. This is a cornerstone of credit risk management in banking industry. Traditionally, statistical models (such as logistic regression model) are used for credit scoring in practice. Despite many advantages of such approach, recent research shows many alternatives that are in some ways superior to those traditional models. This master thesis is focused on introducing ensemble learning models (in particular constructed by using bagging, boosting and stacking algorithms) with various base models (in particular logistic regression, random forest, support vector machines and artificial neural network) as possible alternatives and challengers to traditional statistical models used for credit scoring and compares their advantages and disadvantages. Accuracy and predictive power of those scoring models is examined using standard measures of accuracy and predictive power in credit scoring field (in particular GINI coefficient and LIFT coefficient) on a real world dataset and obtained results are presented. The main result of this comparative study is that...

The imminent emergence of the hydrogen fuel industry has resulted in an urgent mandate for very specific material testing. Although storage of pressurized hydrogen gas is both practical and attainable, demands for increasing storage pressures (currently around 70 MPa) continue to present unexpected material compatibility issues. It is imperative that materials commonly used in gaseous hydrogen service are properly tested for hydrogen embrittlement resistance. To assess material behavior in a pressurized hydrogen environment, procedures were designed to test materials for susceptibility to hydrogen embrittlement. Of particular interest to the field of high-pressure hydrogen in the automotive industry, austenitic stainless steels SUS 316 and 316L were used to validate the test programs. Tests were first performed in 25 MPa helium and hydrogen at room temperature and at -40°C. Tests in a 25 MPa hydrogen atmosphere caused embrittlement in SUS 316, but not in 316L. This indicated that alloys with higher stacking fault energies (316L) are more resistant to hydrogen embrittlement. Decreasing the test temperature caused slight embrittlement in 316L and significantly enhanced it in 316. Alternatively, a second set of specimens was immersed in 70 MPa hydrogen at 100°C until reaching a uniform concentration of absorbed hydrogen. Specimens were then loaded in tension to failure to determine if a bulk saturation of hydrogen provided a similar embrittling effect. Neither material succumbed to the effects of gaseous pre-charging, indicating that the embrittling mechanism requires a constant supply of hydrogen at the material surface rather than having bulk concentration of dissolved hydrogen. Permeation tests were also performed to ensure that hydrogen penetrated the samples and to develop material specific permeation constants. To pave the way for future work, prototype equipment was constructed allowing tensile or fatigue tests to be performed at much higher hydrogen pressures. To determine the effect of pressure on hydrogen embrittlement, additional tests can be performed in hydrogen pressures up to 85 MPa hydrogen. The equipment will also allow for cyclic loading of notched tensile or compact tension specimens for fatigue studies. / Applied Science, Faculty of / Materials Engineering, Department of / Graduate

Recent structural studies evidenced the presence of a recurring well-known interaction between an oxygen atom and an aromatic nucleobase ring in structural motifs of nucleic acids. In particular, this type of interaction is observed between the O4' atom of the (deoxy)ribose moiety and the aromatic nucleobase in Z-DNA molecules and in a variety of structural RNA molecules. In this thesis, we comprehensively examine the hitherto undetected stacking interactions between an oxygen atom of a water (Ow) molecule and the aromatic nucleobase ring, using structural bioinformatics along with quantum mechanics. On the basis of the structural analysis of the high-resolution X-ray structures, we found out that the stacking distance between the Ow atom and the nucleobase plane varies between 3.1 and 4.0 Å. Further, the contact between the Ow-nucleobase plane can be categorized either as a lonepair-π type, where the Ow atom interacts directly with the aromatic surface of the nucleobase, or as an OH-π interaction, where one of the hydrogen atoms of the Ow points towards the nucleobase. Our quantum chemical analysis evidenced that the OH-π interaction is clearly favored in terms of energetics when compared to the lonepair-π, except for the uracil, where the lonepair-π kind of interaction seems to be energetically more stable, as also supported by electrostatic potential map calculations.

Theoretical studies of pi-pi interactions on several dimers of curved polycyclic aromatic systems have been carried out. In the first part, dispersion corrected density functional theory methods (DC-DFT) were used to evaluate the basis set superposition errors (BSSE) in dispersion interactions of the corannulene dimer, and the accuracy of the calculations using DC-DFT methods was compared with high level benchmark calculations. In these calculations, Grimme’s B97D DC-DFT method provided reasonably accurate results with the benchmark calculations. In addition, BSSE obtained with the B97D method along with cc-pVQZ basis set was negligible. Furthermore, a series of calculations were carried out to obtain the pi-pi interaction energy and most stable conformation for the sumanene dimer system. In these calculations, Grimme’s B97D method was used. The potential energy minimum of the sumanene dimer was determined as the concave-convex stacked arrangement with one monomer unit rotated to 60°. The binding energy of the dimer was found to be 19.34 kcal/mol with a 3.72 angstrom distance between two monomer units. Dimers of three different heterosumanenes along with the parent sumanene were also studied. In this set of calculations, two different concave-convex dimer motifs were chosen, eclipsed and staggered (60° rotated). For all the heterosumanenes, as well as the parent sumanene, the staggered conformation is the most stable geometry. The parent sumanene had the highest binding energy. The –NH substituted sumanene produced the second highest binding energy, while the –O analog was the weakest bonded dimer. Finally, dispersion calculations were carried out for the planar aromatic compound of triphenylene. The pi-system of the dimer was distorted by rotating one monomer unit around the principle axis and parallel displacing one monomer unit relative to the other one. Among the rotational dimers, the 39° rotated dimer was the minimum energy conformation. Interaction energy of that dimer was 14.42 kcal/mol with 3.40 angstrom separation between monomers at the B97D/cc-pVQZ level. The parallel displaced minimum energy dimer has a binding energy about 1.0 kcal/mol smaller than the rotational minimum energy geometry.