Includes bibliographical references and index.

Behavioral scientists working in organizations today have access to unprecedented amounts of data. Networked computing and software tools are changing the landscape of organizational research in fundamental ways. Information technology facilitates creation of vast amounts of data. In organizational research, online surveys, interactive interviews, computer-based assessment, and other digital tools have become a preferred medium for collecting self-report and opinion data. Other unobtrusive, nontraditional sources of behavioral observation data are coming into use. Measures of online behavior and databases maintained by corporations and government agencies for other purposes can be a useful source of research data. Concurrent with expanding data availability, analytic capability and processing capacity have improved dramatically. One permutation of this evolution is the recent appearance of computationally intensive methods only recently enabled by spectacular gains in processing speed. These ""brute force"" computational techniques were often developed to solve highly complex problems in the physical sciences and are now migrating into the toolkit of organizational research. Artificial neural networks constitute one class of these powerful new tools. An artificial neural network (ANN) is a statistical model comprised of simple, interconnected processing elements that are configured through iterative exposure to sample data. The most significant departure of neural network analysis from conventional analysis is that neural model development is relatively unconstrained by researcher expectations compared with the defined parameters of anticipated functional relationships inherent to hypothesis testing. Several themes are in this chapter that are discussed in more detail in later chapters. When one or more of the following conditions are present in a research project, neural network analysis may have value in concert with or even in lieu of conventional multivariate analysis: When sample data show high dimensionality, multiple variable types, and complex interaction effects or do not meet parametric assumptions; When evaluation of alternative models is required; When relationships between independent and dependent variables are weak and unexplained variance is large; When the research application supports or requires the use of data-mining procedures; When the theoretical basis of prediction is ambiguous or poorly understood; When operational use of the predictive model requires high fault tolerance; and When conventional modeling is unnecessary or cannot be completed under operational time constraints.

2017"