Fire Prediction: Making Sense of Big Data

file
(AP Photo/Eric Risberg,File)

Big Data still gives fire chiefs and elected officials trepidation and pause when fire departments, academics and data scientists delve into the possibilities and probabilities of predictive analytics. Harvard's Data Smart City Solutions blog recently posted a great piece by Jonathan Jay on using predictive analytics to predict house fires, titled as such: "Can Algorithms Predict House Fires?" This article is a great primer to introduce what some fire departments are doing or have done with Big Data to delve into utilizing their municipalities' metrics to attempt to predict when and where fires may occur.

The muse for the author's piece is the recent Ghost Ship fire in Oakland, California in an artists' residence also used for underground assembly and parties that resulted in one of the greatest losses of life since the Happy Land Social Club Fire in the Bronx in 1990. The Oakland Fire Department was put front-and-center on national news outlets and admitted that their building inspection program failed to identify the structure and its use. Although the blame-game is not over in Oakland, we'd be remiss not to start learning how to prevent such a tragedy in our communities before the dust settles; we know the systemic problems with building inspections and it's up to us to look for the gaps and cracks.


FirefighterNation: Officials Responded to Oakland Warehouse before Fatal Fire

As such, the author of this article uses two seminal data projects conducted by the Atlanta, Georgia and New Orleans, Louisiana fire departments on predicting building fires and determining where to conduct community risk reduction efforts (smoke detector installs), respectively. These two fire departments used the academic community to assist with juxtaposing established statistical models and heuristic ones to test predictive algorithms. The results were amazing and I'll let the article speak for itself, but my M.O. in discussing this great article is where we go from here.

FireRescue Magazine: Big Data in the Fire Service: A Primer

As mentioned above, the trepidation is there as there's always trust, intent and transparency issues when using data to predict, determine and answer questions (some we didn't even know we had to ask). Although I have to admit that we're sobering-up from being 'Data Drunk' with initial outcomes and possibilities, based-upon the unbelievable data points many cities are collecting, we're starting to crest the hill and find the inputs that we actually need to use. Remember, just because we track something and or can access zillions of metrics to incorporate into Random Forest and or Regressive algorithms, doesn't mean they'll add value to the result. In fact, one major U.S. city is backing away from one of the most advanced predictive analytic tools and I'm guessing it's due to the amount of inputs that are being debunked as the weeds tend to get thick, rapidly.

FirefighterNation: Using Data to Strengthen Fire Prevention

I'm very optimistic and excited for what's in store as we hone in on the fire service's Rosetta Stone for Big Data. Now that we've seen demonstrable results on myriad projects across the private and public sector, it's time to roll-up our sleeves and start making sense of the data stuff as it'll be the currency with which we bargain with for more resources at budget time and how we explain why certain buildings get more attention than others when CNN comes to town and puts a microphone in front of our face.

FireRescue Magazine: Jim Crawford, Community Risk Reduction



Current Issue

April 2017
Volume 12, Issue 4
1704fr_C1digonly
Pennwell