2.1 Domain Expertise

One of the critical prerequisites for data mining is something called “domain expertise.” Generally defined, domain expertise implies knowledge and understanding of the essential aspects of a specific field of inquiry. In other words, you need to know your stuff. This is absolutely essential in data mining because so much of the discovery and evaluation process is guided by an intuitive knowledge of what has value, both in terms of input and output, as well as of what makes sense. With a poor understanding of where the information came from and what the results will be used for, the analytical products are unlikely to have much, if any, value. Briefly stated, domain expertise is used to evaluate the inputs, guide the process, and evaluate the end products within the context of value and validity.

Operational personnel think quickly and make rapid decisions because they have to. They also possess extreme confidence in their abilities and knowledge—again, because they have to. To behave any other way would make them inherently unsafe in their profession. If they stopped to ponder all of the possible alternative hypotheses and outcomes like analysts would, they would not last long on the street. They would either be killed by the bad guys or lose the support of their own troops after waiting too long to make a decision.

In most situations, operational personnel know more than anyone else about crime, criminals, crime trends, and patterns, what is “normal,” and what is cause for concern. Given this definition, operational personnel should be natural data miners. Unfortunately, one area where operational personnel seem to lack confidence is in the area of data and analysis. Many have an aversion to statistics and seem to be somewhat intimidated by the whole process. This is really unfortunate because most of them have excellent analytical skills. In many ways, a good investigator is an excellent analyst and a natural data miner. In fact, investigative training and process resembles case-based reasoning in many ways. Investigators typically “understand new [cases] in terms of past ones” that they have investigated.¹

For example, who better understands the limitations of crime and intelligence data than the people responsible for collecting it? Who knows better what the analytical products will be used for and what they should look like? Similarly, who better to distinguish between suspicious data and data that are both valid and reliable? The answer to all three questions is operational personnel. Our sworn partners in the good fight are perfectly suited in many ways to do our jobs as analysts, or at least to partner more closely with us in the analytical process.

2.2 Domain Expertise for Analysts

Analysts that spend all of their time in front of a computer can become so separated from the data and end users that they have little value to the organization. Getting out into the field whenever possible serves at least four separate and important functions. First, fieldwork helps analysts understand the data and where it comes from. This work helps them enhance or, in some cases, begin to acquire their domain expertise. It is very difficult to analyze crime or intelligence data without some understanding of the larger context. Again, it is important to know your subjects/suspects. Certainly, there are situations where it would be dangerous or inappropriate for an analyst to tag along, but periodic ride-alongs, regular attendance at roll call or command staff meetings, and frequent interaction with the organization’s operational personnel provide invaluable education regarding local trends and patterns, as well as insight into historical information and institutional memory. Some of the most teachable moments I have experienced were standing over the victim at a crime scene at two o’clock in the morning or sitting in the back of a sweltering surveillance vehicle in July.

Fieldwork also can be particularly useful in identifying limitations to reliability and validity in the data. Similarly, it is very helpful to understand the operational limitations placed on data collection. In many situations, the operators are the individuals responsible for collecting the data and information. Whether incident reports, surveillance information, informant interviews, or forensic evidence, the data collection task almost always resides with the operational personnel. Unfortunately, this frequently creates a tension between collection of complete, accurate, and reliable information and getting the job done on the street. As nice as it would be to have each and every offense report completed accurately with detailed narrative summaries, good behavioral descriptors, and neat penmanship, this is unlikely in most situations. Given current staffing shortages and workload issues, many sworn personnel are so busy responding to calls and other pressing issues that they end up completing many of their reports during their meal break or at the end of their tour. It would be impossible to completely understand the unique challenges that confront overworked operational personnel; however, until analysts “walk the walk,” the gulf between them and their sworn counterparts will be huge.

Second, by getting out in the field, analysts get a better understanding of what the operational personnel need. For example, the last thing that most sworn folks need or want is more paperwork. Filling out a pile of lengthy field interview reports, particularly if they are cumbersome, duplicative of other reports, and unnecessarily detailed, really pales in importance when faced with multiple pending calls. By getting out into the field, analysts might be able to identify opportunities to streamline reporting and otherwise become part of the solution. This benefits everyone, including the analysts, who are more likely to receive help, guidance, and valuable input from their colleagues in the field when their relationships are enhanced in this manner.

Third, analysts can better target their analytical products by working more closely with the operational units that they support. Getting out from behind the computer increases the give and take. Some of the best research and analysis that I have had the pleasure to be involved with have come from informal conversations with folks who were on the job and said, “Have you ever thought about looking into this?” There is no disgrace in going directly to the end users and working with them to create an analytical product that will meet their needs. It certainly saves time and effort when compared to the all-too-common approach of successive approximations.

Finally, fieldwork helps build the relationship between analysts and operational personnel. There is nothing like standing outside at two o’clock in the morning in a freezing rain to create camaraderie and bonding. Fieldwork helps analysts understand the unique responsibilities, limitations, and time constraints that the operational personnel face in the line of duty. It also sends a strong positive message to the folks working in the field, who generally receive little praise for doing a difficult job under often miserable circumstances.

2.3 Compromise

Clearly, most operators are not about to give up their lives of excitement and adventure to devote the remainder of their professional careers to analyzing data and information, but there are several avenues for collaboration and compromise.

First, viewing the analysis of crime and intelligence data as a partnership offers the unique opportunity to achieve the best of all worlds. As indicated in Figure 2-1, in many agencies data and information arrive at the desk of the analyst, who reviews and analyzes the information and then prepares some sort of analytical product, which is sent up through the command staff and/or out into the field.

Figure 2-1 In the traditional model, analysts prepare reports and other analytical output with little input or feedback from the operational end users.

A revised model that integrates analysis and operations into a seamless, self-perpetuating cycle is outlined in Figure 2-2. By working together, the information comes to the analyst within an operational context. The analyst has some indication regarding where the information came from, its reliability, and its validity, as well as what type of analytical product would be most desirable. The information is then processed and analyzed in a much more meaningful way than if the analyst had been working in an informational void. Similarly, the output, rather than representing some arcane statistical analysis or simple crime count, has operational value that can be appreciated and employed directly by the operators. Certainly, there are situations when it is not possible to share everything with the analyst; however, these situations can be mitigated somewhat with even minimal interaction and guidance on the part of the operational personnel or other end users.

Figure 2-2 By establishing information as a fluid interface between operational and analytical domains, it is analyzed within an operational process. Sworn personnel are able to guide the process, including the nature, structure, and format of analytical output, which increases the likelihood that analytical products will be actionable. In the meantime, the analyst frequently receives better information from the field, while gaining better insight regarding data reliability and validity.

Data mining and predictive analytics, therefore, offer a unique opportunity for analytical and operational personnel to work together in new and exciting ways. By exploiting the intuitive nature of this analytical process, these two groups, with their complementary domain expertise, can more fully utilize existing information resources while creating and guiding novel approaches to enhancing public safety. Although this certainly represents a paradigm shift in how these two relatively diverse professional domains currently function, data mining and predictive analytics afford a unique opportunity to achieve analytical critical mass, taking crime and intelligence analysis into the future.

Think back to one of the scenarios outlined in the Introduction: An agent is debriefing a suspected terrorist in some remote location. The verbal information is automatically recorded, transcribed, and then uploaded to an analytical fusion center a thousand miles away. There, an analyst, using sophisticated text and data mining technology, characterizes and models the results of the interview based on additional information gathered from similar operations throughout the world. Based on this analysis, the agent in the field receives timely information and analysis that enhances his current interview process, while concomitantly increasing the existing intelligence on the operations, practices, and strategies of this particular group. Through this innovative use of expert systems, the intuitive nature of data mining and predictive analytics affords new opportunities for collaboration between analytical and operational personnel that ultimately will enhance our awareness for the war on terrorism, the war on drugs, or the war on crime.

2.4 Analyze Your Own Data

The folks working on the job frequently are in the best position to analyze their own data. Whether the analytical staff, the operational personnel, or a combination of the two, these individuals clearly have the requisite domain expertise to engage in a thoughtful, meaningful analysis of the information and create actionable analytical output that will have value in the field. While it is tempting to pass this task along to outside consultants or statistical gurus, professionals in the public safety arena have a solemn responsibility to those they serve to ensure that they receive the best service possible. Given the domain expertise and experience within the organization, it does not make sense to abdicate this function to someone who might not have the same level of knowledge or understanding just because they have some skills with analytical software. One of the most exciting aspects associated with the newer generation of data mining software applications is that they are intuitive enough to enable even “mere mortals” access to these powerful crime-fighting techniques, which will allow law enforcement and intelligence personnel the opportunity to analyze their own data.