In this prototype, I want to show distribution of call type. Size is count of unit Id, and group by call type. Obviously, the medical incident occupied the most area in this pie chart, so I removed medical incident part to make the percentage of other call type clearly.
Here is the call type distribution without medical incident, and now we can know alarm and structure fire have the most area which sounds like what should fire department should do.
According to other group suggestion, I sort call type by sum number to make data more clear.
According to other group suggestion, I sort call type by sum number to make data more clear.
The graph shows the number of calls to SF fire department from different neighborhoods, categorised by CallTypeGroup. The call type groups are Alarm, Fire, Non life threatening, and Potentially life threatening. The grapg is color encoded with call for better understanding. For some records the categorization is not available. So it is characterised as Null.
The above graph shows the Average response time in minutes, for emergency calls from the neighborhoods of San Francisco. It is color encoded in the increasing order of response time. The darker the color, more the response time.
The purpose of this visualization was to see if the time of year had any effect on the distribution of call types. However as you can see, the data seems pretty unremarkable, so I moved onto more specific data.
Because the distribution of call types seemed to be static throughout the year, I wanted to zoom-in and investigate parameters that could flucuate throughout the day. I chose to examine response time and hospital transport time. Response time was as described on the Data Processing Page.
Obviously, because the data was so large (>300k entries), I needed to use a python script to randomly sample a subset of 2,500 rows of data, import the subset into Tableau, make my calculated fields, and then import the data to rawgraphs.io to make the parallel coordinate.
Click here to look at the script
Click here to see the resulting dataset that is being used.