Master of Science / Department of Electrical and Computer Engineering / Caterina M. Scoglio / West Nile virus (WNV) ---a mosquito-borne arbovirus--- entered the USA through New York City in 1999 and spread to the contiguous USA within three years while transitioning from epidemic outbreaks to endemic transmission. The virus is transmitted by vector competent mosquitoes and maintained in the avian populations. WNV spatial distribution is mainly determined by the movement of residential and migratory avian populations. We developed an individual-level heterogeneous network framework across the USA with the goal of understanding the long-range spatial distribution of WNV. To this end, we proposed three distance dispersal kernels model: 1) exponential ---short-range dispersal, 2) power-law ---long-range dispersal in all directions, and 3) power-law biased by flyway direction ---long-range dispersal only along established migratory routes. To select the appropriate dispersal kernel we used the human case data and adopted a model selection framework based on approximate Bayesian computation with sequential Monte Carlo sampling (ABC-SMC). From estimated parameters, we find that the power-law biased by flyway direction kernel is the best kernel to fit WNV human case data, supporting the hypothesis of long-range WNV transmission is mainly along the migratory bird flyways. Through extensive simulation from 2014 to 2016, we proposed and tested hypothetical mitigation strategies and found that mosquito population reduction in the infected states and neighboring states is potentially cost-effective.

Paleo data-model comparison is the process of comparing output from model simulations of past periods with paleoenvironmental data. It enables us to understand both the paleoclimate mechanism and responses of the earth environment to the climate and to evaluate how models work. This dissertation has two parts that each involve the development and application of approaches for data-model comparisons. In part 1, which is focused on the understanding of both past and future climatic changes/variations, I compare paleoclimate and historical simulations with future climate projections exploiting the fact that climate-model configurations are exactly the same in the paleo and future simulations in the Coupled Model Intercomparison Project Phase 5. In practice, I investigated large-scale temperature responses (land-ocean contrast, high-latitude amplification, and change in temperature seasonality) in paleo and future simulations, found broadly consistent relationships across the climate states, and validated the responses using modern observations and paleoclimate reconstructions. Furthermore, I examined the possibility that a small set of common mechanisms controls the large-scale temperature responses using a simple energy-balance model to decompose the temperature changes shown in warm and cold climate simulations and found that the clear-sky longwave downward radiation is a key control of the robust responses. In part 2, I applied the equilibrium terrestrial biosphere models, BIOME4 and BIOME5 (developed from BIOME4 herein), for reconstructing paleoclimate. I applied inverse modeling through the iterative forward-modeling (IMIFM) approach that uses the North American vegetation data to infer the mid-Holocene (MH, 6000 years ago) and the Last Glacial Maximum (LGM, 21,000 years ago) climates that control vegetation distributions. The IMIFM approach has the potential to provide more accurate quantitative climate estimates from pollen records than statistical approaches. Reconstructed North American MH and LGM climate anomaly patterns are coherent and consistent between variables and between BIOME4 and BIOME5, and these patterns are also consistent with previous data synthesis. This dissertation includes previously published and unpublished coauthored material.

No description available.

The assessment of the performance of learners by means of benchmark experiments is established exercise. In practice, benchmark studies are a tool to compare the performance of several competing algorithms for a certain learning problem. Cross-validation or resampling techniques are commonly used to derive point estimates of the performances which are compared to identify algorithms with good properties. For several benchmarking problems, test procedures taking the variability of those point estimates into account have been suggested. Most of the recently proposed inference procedures are based on special variance estimators for the cross-validated performance. We introduce a theoretical framework for inference problems in benchmark experiments and show that standard statistical test procedures can be used to test for differences in the performances. The theory is based on well defined distributions of performance measures which can be compared with established tests. To demonstrate the usefulness in practice, the theoretical results are applied to benchmark studies in a supervised learning situation based on artificial and real-world data. / Series: Report Series SFB "Adaptive Information Systems and Modelling in Economics and Management Science"

The increasing variety of data mining tools offers a large palette of types and representation formats for predictive models. Managing the models then becomes a big challenge, as well as reusing the models and keeping the consistency of model and data repositories. Sustainable access and quality assessment of these models become limited to researchers. The approach for the Data and Model Governance (DMG) makes easier to process and support complex solutions. In this thesis, contributions are proposed towards ensembles of models with a focus on model representation, comparison and usage. Predictive Toxicology was chosen as an application field to demonstrate the proposed approach to represent predictive models linked to data for DMG. Further analysing methods such as predictive models comparison and predictive models combination for reusing the models from a collection of models were studied. Thus in this thesis, an original structure of the pool of models was proposed to represent predictive toxicology models called Predictive Toxicology Markup Language (PTML). PTML offers a representation scheme for predictive toxicology data and models generated by data mining tools. In this research, the proposed representation offers possibilities to compare models and select the relevant models based on different performance measures using proposed similarity measuring techniques. The relevant models were selected using a proposed cost function which is a composite of performance measures such as Accuracy (Acc), False Negative Rate (FNR) and False Positive Rate (FPR). The cost function will ensure that only quality models be selected as the candidate models for an ensemble. The proposed algorithm for optimisation and combination of Acc, FNR and FPR of ensemble models using double fault measure as the diversity measure improves Acc between 0.01 to 0.30 for all toxicology data sets compared to other ensemble methods such as Bagging, Stacking, Bayes and Boosting. The highest improvements for Acc were for data sets Bee (0.30), Oral Quail (0.13) and Daphnia (0.10). A small improvement (of about 0.01) in Acc was achieved for Dietary Quail and Trout. Important results by combining all the three performance measures are also related to reducing the distance between FNR and FPR for Bee, Daphnia, Oral Quail and Trout data sets for about 0.17 to 0.28. For Dietary Quail data set the improvement was about 0.01 though, but this data set is well known as a difficult learning exercise. For five UCI data sets tested, similar results were achieved with Acc improvement between 0.10 to 0.11, closing more the gaps between FNR and FPR. As a conclusion, the results show that by combining performance measures (Acc, FNR and FPR), as proposed within this thesis, the Acc increased and the distance between FNR and FPR decreased.

Interpreting particular climatic drivers of local and regional vegetation change from paleoecological records is complex. I explicitly simulated vegetation change from the late-Glacial period to the present on the Olympic Peninsula, WA and made formal comparisons to pollen records. A temporally continuous paleoclimate scenario drove the process-based vegetation model, LPJ-GUESS. Nine tree species and a grass type were parameterized, with special attention to species requirements for establishment as limited by snowpack. Simulations produced realistic present-day species composition in five forest zones and captured late-Glacial to late Holocene transitions in forest communities. Early Holocene fire-adapted communities were not simulated well by LPJ-GUESS. Scenarios with varying amounts of snow relative to rain showed the influence of snowpack on key bioclimatic variables and on species composition at a subalpine location. This study affirms the importance of exploring climate change with methods that consider species interactions, transient dynamics, and functional components of the climate.

The increasing variety of data mining tools offers a large palette of types and representation formats for predictive models. Managing the models then becomes a big challenge, as well as reusing the models and keeping the consistency of model and data repositories. Sustainable access and quality assessment of these models become limited to researchers. The approach for the Data and Model Governance (DMG) makes easier to process and support complex solutions. In this thesis, contributions are proposed towards ensembles of models with a focus on model representation, comparison and usage. Predictive Toxicology was chosen as an application field to demonstrate the proposed approach to represent predictive models linked to data for DMG. Further analysing methods such as predictive models comparison and predictive models combination for reusing the models from a collection of models were studied. Thus in this thesis, an original structure of the pool of models was proposed to represent predictive toxicology models called Predictive Toxicology Markup Language (PTML). PTML offers a representation scheme for predictive toxicology data and models generated by data mining tools. In this research, the proposed representation offers possibilities to compare models and select the relevant models based on different performance measures using proposed similarity measuring techniques. The relevant models were selected using a proposed cost function which is a composite of performance measures such as Accuracy (Acc), False Negative Rate (FNR) and False Positive Rate (FPR). The cost function will ensure that only quality models be selected as the candidate models for an ensemble. The proposed algorithm for optimisation and combination of Acc, FNR and FPR of ensemble models using double fault measure as the diversity measure improves Acc between 0.01 to 0.30 for all toxicology data sets compared to other ensemble methods such as Bagging, Stacking, Bayes and Boosting. The highest improvements for Acc were for data sets Bee (0.30), Oral Quail (0.13) and Daphnia (0.10). A small improvement (of about 0.01) in Acc was achieved for Dietary Quail and Trout. Important results by combining all the three performance measures are also related to reducing the distance between FNR and FPR for Bee, Daphnia, Oral Quail and Trout data sets for about 0.17 to 0.28. For Dietary Quail data set the improvement was about 0.01 though, but this data set is well known as a difficult learning exercise. For five UCI data sets tested, similar results were achieved with Acc improvement between 0.10 to 0.11, closing more the gaps between FNR and FPR. As a conclusion, the results show that by combining performance measures (Acc, FNR and FPR), as proposed within this thesis, the Acc increased and the distance between FNR and FPR decreased.

No description available.

No description available.

No description available.