The Springfield-Crime dataset consists of 500,000 police reports filed in Springfield, Missouri between January, 2004 and December, 2016. A second Data-set is provided that includes month by month incident totals. A series of Jupyter Notebooks are included to show analysis.
Data was collected from the now defunct website SpringfieldCrimeAlert using the Pandas read_html() method. A sample script, Scraper.py is provided to show how data was collected.
The data cannot be verified as complete as the collection methods for SpringfieldCrimeAlert(assumed to be a third party and not the Springfield Police Department) are not known. There is a strong case that reports are collected through the police calls search page as descripitions and "approximate location" addresses are identical. This leaves concern that reports are being missed that are filed online. However, its also unclear how many reports are filed online rather than over the phone, and may be statistically insignificant to begin with.
Given that 12 years of data are presented in the dataset the user must be careful in analyzing data without knowing what changes in reporting methods.
For example: taken on its own the following data suggests a decline in non-injury vehicle accidents beginning in 2014
However, there is a second report type. past non-injury vehicle accidents, look what happens when you overlay this graph with the former.
The data suggests a change in reporting procedures in which more non injury vehicle accidents are labeled as past rather than not, however this has not been verified.