Thesis (Ph. D.)--Oregon State University, 2010. / Printout. Includes bibliographical references (leaves 45-50). Also available on the World Wide Web.

This thesis mainly discusses a lexical category known by the name “classifier” or “measure” in Mandarin Chinese. According to Chinese descriptive grammars, which are followed in this thesis, there are at least two types of classifiers, nominal ones (classifiers for nouns) and verbal/ adverbial ones (classifiers for verbs). Within the category of classifiers, there is a subclass that is made of body part terms. Some of them are used for nouns (i.e., NCLs), some of them are used for verbs (i.e., VCLs), and many for both. The correlations between NCLs and VCLs are discussed in this thesis, especially the idea of how both types of classifier can provide delimitation to events. The semantic properties of VCLs and NCLs are explored in this thesis. The [Num+ CL+ N] construction is considered to be a NCLP (nominal classifier phrase) in this thesis and the [V+ [Num+CL]] construction is considered to be a VCLP. I will discuss VCLPs with respect to transitivity of verbs.

The diagnostic evaluation of ever smokers with pulmonary nodules represents a growing clinical challenge due to the implementation of lung cancer screening. The high false-positive rate of screening frequently results in the use of unnecessary invasive procedures in patients who are ultimately diagnosed as benign, clearly highlighting the need for additional diagnostic approaches. We previously derived and validated a bronchial epithelial gene-expression biomarker to detect lung cancer in ever smokers. However, bronchoscopy is not always chosen as a diagnostic modality. Given that bronchial and nasal epithelial gene-expression are similarly altered by cigarette smoke exposure, we sought to determine if cancer-associated gene-expression might also be detectable in the more readily accessible nasal epithelium. Nasal epithelial brushings were prospectively collected from ever smokers undergoing diagnostic evaluation for lung cancer in the AEGIS-1 (n=375) and AEGIS-2 (n=130) clinical trials and gene-expression profiled using microarrays. The computational framework used to discover biomarkers in these data was formalized and implemented in an open-source R-package. We identified 535 genes in the nasal epithelium of AEGIS-1 patients whose expression was associated with lung cancer status. Using matched bronchial gene-expression data from a subset of these patients, we found significantly concordant cancer-associated gene-expression alterations between the two airway sites. A nasal lung cancer classifier derived in the AEGIS-1 cohort that combined clinical factors and nasal gene-expression had significantly higher AUC (0.81) and sensitivity (0.91) than the clinical-factor model alone in independent samples from the AEGIS-2 cohort. These results support that the airway epithelial field of lung cancer-associated injury extends to the nose and demonstrates the potential of using nasal gene-expression as a non-invasive biomarker for lung cancer detection. The framework for deriving this biomarker was generalized and implemented in an open-source R-package. The package provides a computational pipeline to compare biomarker development strategies using microarray data. The results from this pipeline can be used to highlight the optimal model development parameters for a given dataset leading to more robust and accurate models. This package provides the community with a novel and powerful tool to facilitate biomarker discovery in microarray data.

This study examines the relationship between language and cognition with a focus on Chinese numeral classifiers (CNCs). NCs are ideally suited to exploring the link between language and semantic categorization, as classifier selection depends on the physical attributes of the associated noun (e.g., Mandarin zhi is used for long and rigid objects and tiao for long and flexible objects). Previous studies on numeral classifiers have addressed the language-cognition link by comparing the cognitive performance of monolingual as well as bilingual speakers of different languages (Lucy, 1992; Saalbach & Imai, 2005; Gao & Malt, 2009). In contrast, the present study sought to address the cognitive effects of numeral classifiers via a training study that investigated whether exposure to CNCs influenced Native-English speakers’ object categorization preferences, inhibitory control and memory retrieval. The participants of this study were 99 Native-English speaking College students. They were randomly assigned to an experimental group, which received training on four commonly used CNCs during the initial phase of the experiment, or a control group, which did not receive similar treatment during the initial phase. After the initial phase, the experimental group and the control group were assessed on a Forced Choice Task, a Go/No-Go Task and a Memory Task. A Mixed-design ANOVA indicated that the experimental group displayed a preference for objects sharing the same classifier in the Forced Choice Task and the Go/ No-Go Task (i.e. Go trials) when compared to the controls. The effect of exposure to numeral classifiers on inhibitory control was supported with a significantly lower false alarm rate (in the No-Go trials) for the experimental group. However, no group differences were observed in the results of the analysis of the participants’ median reaction times in the Go/No-Go tasks. Similarly, the differences between the two groups’ scores on the Memory Task was not found to be significant. The results of the study indicated that exposure to CNCs influenced Native-English speakers’ categorization. The results also revealed partial support for the influence of exposure to CNCs on inhibitory processing, but not in the case of object clustering.

Access technologies often deal with the classification of several physiological and biomechanical signals. In most previous studies involving access technologies, a single classifier has been trained. Despite reported success of these single classifiers, classification accuracies are often below clinically viable levels. One approach to improve upon the performance of these classifiers is to utilize the state of- the-art multiple classifier systems (MCS). Because MCS invoke more than one classifier, more information can be exploited from the signals, potentially leading to higher classification performance than that achievable with single classifiers. Moreover, by decreasing the feature space dimensionality of each classifier, the speed of the system can be increased. MCSs may combine classifiers on three levels: abstract, rank, or measurement level. Among them, abstract-level MCSs have been the most widely applied in the literature given the flexibility of the abstract level output, i.e., class labels may be derived from any type of classifier and outputs from multiple classifiers, each designed within a different context, can be easily combined. In this thesis, we develop two new abstract-level MCSs based on "reputation" values of individual classifiers: the static reputation-based algorithm (SRB) and the dynamic reputation-based algorithm (DRB). In SRB, each individual classifier is applied to a “validation set”, which is disjoint from training and test sets, to estimate its reputation value. Then, each individual classifier is assigned a weight proportional to its reputation value. Finally, the total decision of the classification system is computed using Bayes rule. We have applied this method to the problem of dysphagia detection in adults with neurogenic swallowing difficulties. The aim was to discriminate between safe and unsafe swallows. The weighted classification accuracy exceeded 85% and, because of its high sensitivity, the SRB approach was deemed suitable for screening purposes. In the next step of this dissertation, I analyzed the SRB algorithm mathematically and examined its asymptotic behavior. Specifically, I contrasted the SRB performance against that of majority voting, the benchmark abstract-level MCS, in the presence of different types of noise. In the second phase of this thesis, I exploited the idea of the Dirichlet reputation system to develop a new MCS method, the dynamic reputation-based algorithm, which is suitable for the classification of non-stationary signals. In this method, the reputation of each classifier is updated dynamically whenever a new sample is classified. At any point in time, a classifier’s reputation reflects the classifier’s performance on both the validation and the test sets. Therefore, the effect of random high-performance of weak classifiers is appropriately moderated and likewise, the effect of a poorly performing individual classifier is mitigated as its reputation value, and hence overall influence on the final decision is diminished. We applied DRB to the challenging problem of discerning physiological responses from nonverbal youth with severe disabilities. The promising experimental results encourage further development of reputation-based multi-classifier systems in the domain of access technology research.

Access technologies often deal with the classification of several physiological and biomechanical signals. In most previous studies involving access technologies, a single classifier has been trained. Despite reported success of these single classifiers, classification accuracies are often below clinically viable levels. One approach to improve upon the performance of these classifiers is to utilize the state of- the-art multiple classifier systems (MCS). Because MCS invoke more than one classifier, more information can be exploited from the signals, potentially leading to higher classification performance than that achievable with single classifiers. Moreover, by decreasing the feature space dimensionality of each classifier, the speed of the system can be increased. MCSs may combine classifiers on three levels: abstract, rank, or measurement level. Among them, abstract-level MCSs have been the most widely applied in the literature given the flexibility of the abstract level output, i.e., class labels may be derived from any type of classifier and outputs from multiple classifiers, each designed within a different context, can be easily combined. In this thesis, we develop two new abstract-level MCSs based on "reputation" values of individual classifiers: the static reputation-based algorithm (SRB) and the dynamic reputation-based algorithm (DRB). In SRB, each individual classifier is applied to a “validation set”, which is disjoint from training and test sets, to estimate its reputation value. Then, each individual classifier is assigned a weight proportional to its reputation value. Finally, the total decision of the classification system is computed using Bayes rule. We have applied this method to the problem of dysphagia detection in adults with neurogenic swallowing difficulties. The aim was to discriminate between safe and unsafe swallows. The weighted classification accuracy exceeded 85% and, because of its high sensitivity, the SRB approach was deemed suitable for screening purposes. In the next step of this dissertation, I analyzed the SRB algorithm mathematically and examined its asymptotic behavior. Specifically, I contrasted the SRB performance against that of majority voting, the benchmark abstract-level MCS, in the presence of different types of noise. In the second phase of this thesis, I exploited the idea of the Dirichlet reputation system to develop a new MCS method, the dynamic reputation-based algorithm, which is suitable for the classification of non-stationary signals. In this method, the reputation of each classifier is updated dynamically whenever a new sample is classified. At any point in time, a classifier’s reputation reflects the classifier’s performance on both the validation and the test sets. Therefore, the effect of random high-performance of weak classifiers is appropriately moderated and likewise, the effect of a poorly performing individual classifier is mitigated as its reputation value, and hence overall influence on the final decision is diminished. We applied DRB to the challenging problem of discerning physiological responses from nonverbal youth with severe disabilities. The promising experimental results encourage further development of reputation-based multi-classifier systems in the domain of access technology research.

Background: Over the last decade, there has been a significant rise in the life insurance industry. Every life insurance application is associated with some level ofrisk, which determines the premium they charge. The process of evaluating this levelof risk for a life insurance application is time-consuming. In the present scenario, it is hard for the insurance industry to process millions of life insurance applications.One potential approach is to involve machine learning to establish a framework forevaluating the level of risk associated with a life insurance application. Objectives: The aim of this thesis is to perform two comparison studies. The firststudy aims to compare the accuracy of the logistic regression classifier, decision tree classifier, random forest classifier and linear support vector classifier for evaluatingthe level of risk associated with a life insurance application. The second study aimsto identify the impact of changes in the dataset over the accuracy of these selected classification models. Methods: The chosen approach was an experimentation methodology to attain theaim of the thesis and address its research questions. The experimentation involvedcomparing four ML algorithms, namely the LRC, DTC, RFC and Linear SVC. These algorithms were trained, validated and tested on two datasets. A new dataset wascreated by replacing the "BMI" variable with the "Life Expectancy" variable. Thefour selected ML algorithms were compared based on their performance metrics,which included accuracy, precision, recall and f1-score. Results: Among the four selected machine learning algorithms, random forest classifier attained higher accuracy with 53.79% and 52.80% on unmodified and modifieddatasets respectively. Hence, it was the most accurate algorithm for predicting risklevel in life insurance application. The second best algorithm was decision tree classifier with 51.12% and 50.79% on unmodified and modified datasets. The selectedmodels attained higher accuracies when they are trained, validated and tested withunmodified dataset. Conclusions: The random forest classifier scored high accuracy among the fourselected algorithms on both unmodified dataset and modified datasets. The selected models attained higher accuracies when they are trained, validated and tested with unmodified compared to modified dataset. Therefore, the unmodified dataset is more suitable for predicting risk level in life insurance application.

No description available.

No description available.

Given a limited number of samples for classification, several issues arise with respect to design, performance and analysis of classifiers. This is especially so in the case of microarray-based classification. In this paper, we use a complexity measure based mixture model to study classifier performance for small sample problems. The motivation behind such a study is to determine the conditions under which a certain class of classifiers is suitable for classification, subject to the constraint of a limited number of samples being available. Classifier study in terms of the VC dimension of a learning machine is also discussed.