Content area
Abstract
Potential employers considering new locations for production and service facilities typically require advance knowledge of the wages they will be expected to offer in various occupational categories. The Department of Labor and Industrial Relations in the state of Missouri is often contacted by organizations requesting such information. The current wage rate survey approach, initiated in 1988, allows the department to predict an appropriate wage rate for a given occupation in certain counties, adjusted for inflation. A major deficiency of the current approach is its inability to predict wages for unsurveyed counties.
This dissertation describes the development and application of a knowledge-based system (KBS) to improve the wage rate prediction responsiveness and accuracy. Artificial intelligence language (Prolog) and techniques are employed in the design and development process. First, the wage rate prediction task is analyzed and broken down into two subtasks: county similarity determination task and database searching task. For county similarity determination a weighted Manhattan distance measure, which relies on pairwise difference between the attribute values of instances, was used. For the database searching task, the domain expert's knowledge was extracted and implemented within the system.
In this dissertation, (1) methodologies and formalisms developed in AI literature will be addressed and applied through the actual development of a knowledge-based system, (2) the decision making effectiveness and performance of the system will be demonstrated, and (3) the validation process concerning the system's performance will be documented.
