NJIT
A proposed model for comparing the performance of neural networks and statistical approaches in predicting project profits of an asphalt paving company
Abstract
dc:description.abstract<p>It is currently assumed that the performance of a model for predicting project profits of an asphalt paving company varies with an unknown number of parameters. An earlier study indicates that the parameters are not well understood (Berry, 1990). A search of the literature suggests that the available models usually are described by those who develop and use them as universally reliable and valid tools for solving complex problem (Allman, 1989; Bumke, 1988; Clark, 1988; Cohen & Howe, 1988; Fillon, 1989; Kling, 1990; Lawrence, Petterson, & Hartzberg, 1990; Marose, 1990; Pollak, 1988; Reynolds, 1988; Sullivan & Reeve, 1988; Thurber, 1988). Present models such as neural networks and advanced statistical approaches have been promoted as being capable of augmenting, or even supplanting, human decision-making. Yet, little research published to date compares the results of these two different models (Caudill, 1990). Few works published to date establish any criteria for evaluating the results provided by these models.</p><p>It is the author's belief that neural network models must be subjected to the same rigorous quantitative and qualitative evaluation that statistical models have endured.</p><p>This author believes that the importance of a quantitative metric for measuring the relative accuracy, reliability and validity of the results of each model is apparent. What is less apparent are the qualitative factors, including perceived accuracy, perceived reliability and perceived validity of each model, usability issues, and the cognitive styles of those using the models.</p>
Degree
thesis:*- Name thesis:degree_name
- Master of Science in Management - (M.S.)
- Discipline thesis:degree_discipline
- School of Industrial Management
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Giaimo, Josephine
- Contributors dc:contributor
-
- Gordon Kalley
- Carl Wolf
Subjects
dc:subject × 6Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.njit.edu/theses/2678