Academic journal article Child Welfare

Data Mining in Child Welfare

Academic journal article Child Welfare

Data Mining in Child Welfare

Article excerpt

Data mining is the sifting through of voluminous data to extract knowledge for decisionmaking. This article illustrates the context, concepts, processes, techniques, and tools of data mining, using statistical and neural network analyses on a dataset concerning employee turnover. The resulting models and their predictive capability, advantages and disadvantages, and implications for decision support are highlighted.

Child welfare agencies, like other organizations, are in the process of rapid change, primarily due to the growth in the use of information technology. Agencies have been automating records for the last 10 to 15 years, but it is only in recent years that these systems have collected detailed data on agencies, personnel, clients, and services. Although many agencies now have sophisticated information systems, few are using agency data as a key resource to guide decisionmaking. Decision support is especially needed at the worker level, where task complexity, lack of training, and high turnover prevail. Data mining, a new field, entails sifting through voluminous data and records to extract knowledge for decisionmaking. A 1997-98 META Group study found that nearly 80% of companies expected data mining to be a critical success factor in 1999 [Levy 1999].

This article examines the historical and larger context of data mining and describes data mining processes, techniques, and tools. These are illustrated using a child welfare dataset concerning employee turnover that is "mined" using logistic regression and a Bayesian neural network. A discussion of the data mining process, the resulting models, their predictive capability, their advantages and disadvantages, and their implications for decision support concludes the article.

History and Context

Organizations have always been good data accumulators. Employees, however, are often drowning in data, but starved for knowledge [Naisbitt 1982]. Modern computer and information technologies allow organizations to change data into information and knowledge. One modern management task is to create the organizational structure and processes whereby each employee has instant and easy access to the accumulated knowledge of the agency and the capacity to use that knowledge to improve his or her job performance [Ikujiro & Hirotaka 1995]. This collective knowledge or "organizational intelligence" implies that data are a shared resource, with everyone responsible for collecting, publishing, and using them. It also assumes an agency infrastructure that collects, stores, and manipulates data into knowledge for all to use. Thus, data become an interactive, fluid asset through which employees and stakeholders share, learn, improve, and create a more intelligent agency [Schoech 1999]. A primary role of management is transferring agency data and worker expertise into accumulated agency knowledge. Knowledge management involves systematically capturing organizational information and expertise, integrating it, and making it interactively available to employees who are trained to use it in decisionmaking to achieve goals. Knowledge management tasks include data warehousing, data mining, accumulating expertise, information dissemination, and organizational learning.

This article focuses on the data mining task of knowledge management. Data mining can be formally defined as an analytic process designed to explore large amounts of data in search of consistent patterns or systematic relationships between variables, and then to validate the findings by applying the detected patterns to new subsets of data [Statsoft 1999]. Data mining can be distinguished from other forms of research in that with data mining, the dataset is explored without specific hypotheses to test. Data mining does not concern, disregard, or devalue intuition. Intuition falls within the knowledge management task of accumulating expertise.

Data Mining Processes, Tools, and Techniques

Data mining can be best understood by examining its processes, tools, and techniques. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.