Academic journal article Educational Technology & Society

Predicting Adolescent Deviant Behaviors through Data Mining Techniques

Academic journal article Educational Technology & Society

Predicting Adolescent Deviant Behaviors through Data Mining Techniques

Article excerpt

Introduction

In past, research (Freisthler, Byrnes, & Gruenewald, 2009; Soenens, Vansteenkiste, Smits, Lowet, & Goossens, 2007) has pointed out that if deviant behaviors cannot be indentified and corrected as they occur, adolescents might do something uncontrollable or unredeemable later in their lifetimes. Therefore, in order to avoid consuming social resources in future, deviant behaviors had better be detected and corrected as early as possible. For this reason, many researchers strive to study the prevention and amendment of adolescent deviant behaviors to avoid further serious social problems. Further, some researchers also discuss how to help the young devote themselves to meaningful activities and goals.

From the perspective of educators, the deviant behaviors might greatly influence students' learning progresses since some deviant behaviors are directly correlated in learning, such as: negative learning attitude, truancy, and disobedience. Others might result in addictions or bad relationships; therefore students could hardly focus on academic learning. Therefore, preventing adolescents from deviant behaviors could also improve their learning progresses and abilities to acquire knowledge.

As information technology prevails worldwide, electronic databases have been used to store mountains of data with regards to various fields. Information systems are efficient at collecting data as well as transforming data into useful information. There are many IT tools which can integrate data from different sources, so that the cost of information processing can be reduced and resource allocation can be much more efficient. So with regards to supervising students' deviant behaviors, archived counseling records are very helpful for counselors to trace, monitor, and record behaviors, treatments, and outcomes.

In general, the records stored in database can be in the form of structured or unstructured data fields. The structured fields refer to data which can be briefly summarized as values and stored in certain fields. Contrarily, unstructured fields hold data which is difficult to categorize as finite sets of values and therefore stored intact. Comparing to structured fields, unstructured data fields are difficult to understand and be utilized.

In this paper, we aim to propose a framework to analyze the unstructured but crucial text in databases. We term such unstructured data as "memo-type" data throughout this paper. Such memo-type data is usually presented in the form of short paragraphs, but these paragraphs are dissimilar to regular articles. "Memo-type" records exist in many real applications, such as a patients' diagnosis and treatment data, the counseling records kept by psychiatrists, the maintenance records for hardware equipment, and answers from open questions, and working diaries or journals. While it is known that the text stored in memo-type fields is good at providing details, data utilization is regrettably hampered.

Traditionally, in order to retrieve information from memo-type fields, experts need to transform them into structured data according to domain knowledge. However in practice, such methods encounter certain problems. Firstly, the transformation relies heavily on expert intervention, but various experts have different opinions. Secondly, the newly transformed fields designed by domain experts often result in data records that are too sparse to maintain. Hence, to develop data records which are simple in nature but within which major information is stored in memo-type fields, the method stated above is unsuitable.

In summary, due to the characteristics of memo-type records, traditional document mining techniques are untenable, and the mining task for these memo-type fields should be handled with a much more focused search strategy. To the best of our knowledge, no literature focuses on mining associations from "memo-type" records; therefore, in this research, we propose a semi-automatic approach to analyze memo-type records with only little expert interventions. …

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.