NotesFAQContact Us
Search Tips
ERIC Number: ED549964
Record Type: Non-Journal
Publication Date: 2012
Pages: 112
Abstractor: As Provided
Reference Count: N/A
ISBN: 978-1-2672-7108-2
A Crime Analysis Decision Support System for Crime Report Classification and Visualization
Ku, Chih-Hao
ProQuest LLC, Ph.D. Dissertation, The Claremont Graduate University
Today's Internet-based crime reporting systems make timely and anonymous crime reporting possible. However, these reports also result in a rapidly growing set of unstructured text files. Complicating the problem is that the information has not been filtered or guided in a detective-led interview resulting in much irrelevant information. To facilitate the crime analysis process, a decision support system was developed. The system has been iteratively developed and tested. Its main contributions are 1) the entity extraction algorithm that extracts crime-related entities, e.g., weapon, vehicle, and clothing, 2) similarity scoring algorithm that measures similarities between entities and documents, and 3) information visualization that displays crime reports and their relationships. The "information extraction algorithm" was evaluated with police and witness crime reports. Overall, high precision and recall were achieved in extracted entities for police and witness crime reports. The "similarity scoring algorithm" has two main components: the entity and "document similarity algorithm" and each was evaluated separately. The entity similarity algorithm was evaluated by comparing it to two WordNet-based algorithms with fifteen human raters. Overall, the highest correlation coefficient was obtained with human raters for the proposed algorithm. Two studies were conducted to evaluate document similarity algorithm by classifying documents as discussing the same or a different crime and compare with a crime analyst's performance. The first study used small datasets of crime reports discussing a few types of crimes, while the second used big datasets, discussing more different crimes. Overall, the crime analyst's grouping accuracy decreased slightly, while the system's classification accuracy increased significantly when the number of crime reports and of different crimes increased. To measure usefulness of the visualization, a case study was conducted with a crime analyst. The crime analyst discovered the same number of crime reports correctly among the given reports using either the paper-based or system-based approach. However, less time was spent on the tasks using the proposed system compared to the traditional paper-based approach. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page:]
ProQuest LLC. 789 East Eisenhower Parkway, P.O. Box 1346, Ann Arbor, MI 48106. Tel: 800-521-0600; Web site:
Publication Type: Dissertations/Theses - Doctoral Dissertations
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A