Graduate Theses & Dissertations

Utilizing Class-Specific Thresholds Discovered by Outlier Detection
We investigated if the performance of selected supervised machine-learning techniques could be improved by combining univariate outlier-detection techniques and machine-learning methods. We developed a framework to discover class-specific thresholds in class probability estimates using univariate outlier detection and proposed two novel techniques to utilize these class-specific thresholds. These proposed techniques were applied to various data sets and the results were evaluated. Our experimental results suggest that some of our techniques may improve recall in the base learner. Additional results suggest that one technique may produce higher accuracy and precision than AdaBoost.M1, while another may produce higher recall. Finally, our results suggest that we can achieve higher accuracy, precision, or recall when AdaBoost.M1 fails to produce higher metric values than the base learner. Author Keywords: AdaBoost, Boosting, Classification, Class-Specific Thresholds, Machine Learning, Outliers
Smote and Performance Measures for Machine Learning Applied to Real-Time Bidding
In the context of Real-Time Bidding (RTB) the machine learning problems of imbalanced classes and model selection are investigated. Synthetic Minority Oversampling Technique (SMOTE) is commonly used to combat imbalanced classes but a shortcoming is identified. Use of a distance threshold is identified as a solution and testing in a live RTB environment shows significant improvement. For model selection, the statistical measure Critical Success Index (CSI) is modified to add emphasis on recall. This new measure (CSI-R) is empirically compared with other measures such as accuracy, lift, efficiency, true skill score, Heidke's skill score and Gilbert's skill score. In all cases CSI-R is shown to provide better application to the RTB industry. Author Keywords: imbalanced classes, machine learning, online advertising, performance measures, real-time bidding, SMOTE

Search Our Digital Collections

Query

Enabled Filters

  • (-) ≠ Furgal
  • (-) = Applied Modeling and Quantitative Methods
  • (-) ≠ Avusuglo
  • (-) = Information science

Filter Results

Date

2011 - 2021
(decades)
Specify date range: Show
Format: 2021/04/11

Degree

Subject (Topic)