Publication Date

2020

Document Type

Thesis

Committee Members

Subhashini Ganapathy, Ph.D. (Advisor); Michael Raymer, Ph.D. (Committee Member); Sherif Elbasiouny, Ph.D. (Committee Member)

Degree Name

Master of Science in Industrial and Human Factors Engineering (MSIHE)

Abstract

The effectiveness of representing higher dimensional data on a two-dimensional visualization was required to be studied in the development of a novel data imputation method, Continuous Imputation With Association Rules (CIWAR). When the CIWAR method is used to impute missing data, the method generates additional metadata that increases each imputed data point's dimensionality. Potential use cases for CIWAR include situations where imputed data would be analyzed by individuals with little or no data analytic experience or situations where imputed data would be used to aid high-stress time-critical human decision processes. A study was conducted to assess the effect of the addition of entropy data on trust associated with the imputed data points and the ability to identify trends in imputed data from the CIWAR data visualization. The study examined the responses of two treatment groups (experts and novice), an expert group that had more statistical analysis experience and a novice group with little to no statistical analysis experience. The CIWAR method was capable of imputing data with a 0.043 Normalized Root Mean Squared Error (NRMSE) and was compared to the K-Nearest-Neighbor imputation method that produced a NRMSE of .338. The study showed a correlation between the participant's level of trust and expected level of trust when entropy data is associated with the imputed data.

Page Count

381

Year Degree Awarded

2020


Share

COinS