If we get an emergency call from a specific area at a specific time, can we approximate what emergency it could be?
If we roughly equate the proportional frequencies of emergency call types in January to likelihood, we can approximate the likeliest type of emergency in a given area in a given time of day. Technical details of how this was done can be found in the Jupyter Notebook above, but essentially we filter the data to the desired parameters of location and time without making it too specific, and then generalize off of that observed frequency.
For each of the three applets below, simply choose an area and a time from the drop downs, and click "Predict" to observe what our knowledge of the month of January tells us about the top three most likely emergencies to occur.
Predicting Call Type with Zipcode and Day Portion of Call Received
Day portions are groupings of similar hours of the day- 12 AM to 7 AM is late night hours, 8 AM to 3 PM is morning to late afternoon, and 4 PM to 11 PM is late evening to night hours.
Zipcode:
Day Portion of Call:
Predicting Call Type with Zipcode and Specific Hour the Call was Received
Here we try a second prediction technique in which we select the precise hour at which the call was received, instead of the broader day portion.
Zipcode:
Precise Hour Call was Received (0 (midnight) - 23):
Predicting Call Type with Station Area and Day Portion of Call Received
Here we try a third technique in which we select more granular location; station area rather than the slightly larger zipcodes. The time option remains in day portions to avoid the issue of too few samples in observed data.
Station Area (1-51):
Day Portion of Call:
Predicting Call Type with Station Area and Specific Hour the Call was Received
Here we try a fourth technique in which we select more granular location; station area rather than zipcodes as well as more granular time, specific hour rather than day portion. Fair warning: Due to high specificity, this may return Unknown for some cases, and generalizes much less than the predictors above.
Station Area (1-51):
Precise Hour Call was Received (0 (midnight) - 23):
Predicting Unit Type with Zipcode and Specific Hour the Call was Received
Now we attempt to predict the type of response unit that will be needed for an emergency- Medic, Engine, Truck, etc.- given a zipcode and the hour an emergency call is received.
Zipcode:
Precise Hour Call was Received (0 (midnight) - 23):