So where does the data take me?
Reflections about the last week
I just finished a week at NENA’s Standard and Best Practices. I had a wonderful time with my presentations on developing staffing models, data management, and using data better in call processing. I met a lot of new friends and I see a lot of opportunities ahead of me to make a real difference at the intersection of public safety and data science.
NENA’s SBP
I participated in seven presentations. I will upload slide decks and notes later for people who are interested in learning what I was talking about in detail.
Staffing
Three of these were in a block discussing the need for a new staffing model recommendation in 911 centres. Currently, NENA’s Staffing Recommendations are over 21 years old and rely heavily on the Erlang-C model for staffing recommendations. There has been a tonne of work in queuing theory since the development of this formula. And, as referenced in the linked paper, there are legitimate questions about how appropriate the Erlang model family is for regular call centres, let alone 911 centres.
One of the central problems is that Erlang-C models assume there is only one queue that delivers all of the customers to the servers. In most modern centres, there are at least 4 ingression points to the server set. We have dedicated 9-1-1 trunk lines, typically most primary PSAPs also answer some non-emergency lines, and most centres also take 9-1-1 SMS messages. Finally, all centres have to be able to receive and process TTD-TTY calls. Many larger centers may also process traffic cameras, ASAP2PSAP generated alarm calls, or other monitoring services like ShotSpotter. Dispatchers also have radio traffic from the units that they monitor which require their attention. So, there are numerous queues to access the servers, the servers are all multi-skilled servers, and the different queues have different priorities for servers depending on skill sets and responsibilities. Each of these adds levels of complexity that could confound an Erlang-C model.
There are several different methods that could come into play, including Markovian models, Game Theory, and Priority Queuing Models. Each of these can be explored and addressed.
I remain convinced that we have enough brain power in NENA that our working group can tackle this issue and come up with a queuing model that will work for the industry.
EIDO and NG9-1-1
Two of the other presentations centered around the Emergency Incident Data Object. This is how, we propose, to move information between different systems and entities in public safety. It’s in a JSON object that leverages NIEM to define the objects passed between entities. My presentations dicussed the JSON object itself and the conveyance mechanisms by which the JSON objects are moved between entities.
I believe in the use of the EIDO as we continue to roll out NG9-1-1 services across the country. We have done a lot of good work throughout the years to create something that can be useful in ensuring data gets to where it needs to go and can be processed efficiently.
Data Usage
The other presentations in which I participated were the Call Processing working group where I am the ‘data SME’ and the Future Think committee’s Data Management and Presentation work. That was a fun panel discussion with a lot of great audience participation. There is a general agreement that we need to use more of our data and use it more effectively.
I am heartened, as a data scientist working in public safety, to hear my colleagues in public safety acknowledging the need to use data analytics, regression modeling, predictive analytics, and forecasting to improve things for telecommunicators.
There is also a movement to start considering a shared data repository for research and analytical purposes. I beieve that will be an excellent idea. If we can get several PSAPs together and build out a research database, we could put it in front of other researchers and see where it leades us.
Editorial Note
I changed the link for NIEM to go to the About page to give readers a better explanaiton of what NIEM is and why it’s important. I also linked to a page describing JSON in better terms. Finally, my editor found a spelling error that was also corrected.