6.0 Conclusion - WayneWang86/UDistrict-Real-Time-Fire-Response GitHub Wiki

Strengths:

The strengths of our project are that our data visualizations are easy to read and modify with multiple different widgets that facilitate easy data wrangling. Our visualizations take in well over a million data points that span many years, allowing us to identify patterns and trends in the data. By transforming the data into an interactive map we are able to restrict observations to specific time periods and specific areas even within the U-District for more in-depth analysis. Looking at more than one facet of the data allows us to provide a more complete picture of the issue. Specifically, we investigated the type, location, and time of the incident.

Weaknesses:

• Our project has definite weaknesses; it doesn't see people, and geographically the scope is small. The data set, and in turn, our visualizations lack information about the people at any given incident. The numbers of people involved, lives lost, and other human-focused metrics are not present, so assessing the "importance" of any incident isn't possible with this data alone. Low-risk incidents are represented on our map just as well as the most deadly ones. Fire departments cannot draw fully informed conclusions from our data set, but they can see where and when reports happen. We attempted to somewhat fix this by allowing our histogram visualization to sort by category, but the full picture is still unclear.

• Another shortcoming is that despite having many recorded data points, we exclude many that are not relevant to the U-District. Perhaps the trends we found would be wildly different in different parts of the country, or even within just the city of Seattle.

• According to Sampson, except for the 911 fire calls with real emergencies, there are all other types of false 911 calls, such as Non Emergency calls, prank calls, and even Lonely Complainant calls. These calls do not actually pose any threats or emergencies. As a result, many fire calls recorded in the database might not be a 911 fire call related to real danger or emergency. This could make our analysis result less reliable.

Future Works:

In a follow-up project we should try to include the data on the "people" aspect. One thing we missed out on was the human aspect of our data set. Data like the number of people injured or killed, damage to property, and the scale of the incident would enable us to provide more insightful figures. We would love to analyze the trends and relationships between types of reported incident and the number of people involved/overall damage. Using metrics like these, we could help the fire department to determine which incidents receive higher priority when resources are scarce through a data-driven approach. (This research is not aiming to weigh the value of life among different types of incidents, but to help the U-District Fire Department and other fire departments to deploy their resources more efficiently.)