4.11 Data Mining Agents
Investigative data mining represents a powerful new approach to criminal monitoring detection and alert dissemination. In private industry, data mining has primarily been applied to very large corporate databases for such applications as identifying potential customers. While data mining was originally conceived of as a way of extracting hidden associations from large databases, when coupled with agent technology it can be used to monitor events, extract important information via the Internet, intranets, and other proprietary networks, discover new patterns, assemble profiles, and deliver alerts to military, medical, law enforcement, and intelligence agency personnel. Using sensors, agents can work in tandem with other systems to analyze collected data and then issue real-time alerts to systems or personnel via the Internet, to proprietary networks, even to wireless devices.
For example, IBM developed an agent to work with its Intelligent Data Miner suite. The system consists of five agents:
A user interface agent that provides a Web interface for users to interact with the data miner and help them perform data mining analysis and display results
A coordinator agent that is responsible for delegating and managing various tasks that need to be performed for problem solving
A data-set agent that is responsible for keeping track of what data is stored in which data mart or data warehouse and actively maintaining metadata information
A data mining agent that executes the user-defined algorithms, performs on-line analytical processing (OLAP) analysis, and communicates the results to users or other agents
A visualization agent that allows for ad hoc and predefined reporting capabilities and a wide array of graphical reports
As we shall see in some of the following chapters, other data mining software tools are already incorporating agents into their products to assist the user in minimizing the effort of extracting, preparing, modeling, and delivering the results of their analyses. In the next chapter, we will explore how agent technology can be used with text mining technologies to monitor and retrieve specific information via the Web and other networks for criminal and terrorist detection and case development.
4.12 Agents Tools
Agent software is readily available from various sources, including the Web, from such sites as agentland.com (see Figure 4.3).
Figure 4.3: Agentland.com provides agent software for downloading.
Agents that perform different functions can also be downloaded from agentland (see Figure 4.4). There are information retrieval agents, monitoring agents, and development kits for constructing agents.
Figure 4.4: A menu of development agent software available.
To illustrate how some of these agents work, we will use InfoGIST, an information retrieval agent to search and aggregate a set of Web pages on the basis of keywords we provide: the search will be for "investigative data mining." The agent is instructed to retrieve all pages with all of these words in them and to prioritize them in a list (see Figure 4.5).
Figure 4.5: The completed agent form.
Figure 4.6 shows the results of the agent's search.
Figure 4.6: A list is generated with scores of relevance associated with them.
Using an agent, as opposed to a search engine, has the advantage that all of these links can be viewed at any time. The links are also scored for relevance. This particular agent also assigns a grade to each item it retrieves, with the most relevant earning an A. Portions of the documents are shown on the right screen, with links and the search engine used presented in the upper right window. Analysts can use agents routinely to search for relationships between new data, update old data, perform specific calculations, etc.