It’s not news that there has been a nation wide hike in crimes across the United States, including Los Angeles. NPR episode on LAPD. Inequality and crimes against fellow humans are disheartening and could often seem impossible to resolve. Open data provides one resource for viewing impossible problems and collaborating on solutions. In this analysis, LA city domestic violence counts are viewed in time series and compared by areas in the city.
Summary from the review of LA domestic violence time series:
- What programs could be extended to mitigate the impact of family violence when school-age children are home from school? Below reviews show DV crimes increase in the summer time and before the holidays and New Year. There is also more financial burdens on families related to food, material goods (holiday presents), and activities. Summer activities that extend the school year for children but not increase the burden on teaching and testing are good options for families and the community.
- Spikes in DV during special events, with the highest impact on New Years or January 1st. Does alcohol, substances, or celebration affect this trend? Can the community offer substance free events and gatherings.
- What environmental attributes of each neighborhood affects the level of violence in them? Is there more adverse advertising in communities with more DV, such as negative depictions of women, alcohol and liquor advertisements, and etc.
- What personal attributes contribute to an increase of domestic violence? Another data set is required for this type of analysis. This is a supervised learning problem.
- What other factor or data to connect with DV occurrences? The relatively weak correlation (but still a slight correlation) between domestic violence and time shows that the problem permeates the community at all times. Other causes and correlations should be explored.
- Meta suggestions: Help families with changes in scheduling from school, work, and holidays
I reviewed patterns of crimes with relationship to time from LA city crimes open data set. Crime data was available for 2013 and 2014 datasets from the City of Los Angeles Open Data. Each of the unique 153 crime categories in the data set have their own set of “rules”, influencing factors, and impact on the communities.
[table “” not found /]
Domestic violence related crimes (“DV”) are crimes described as containing “Spousal (cohab) abuse.” The DV crime counts by area shows where dv crimes are reporting to the police most often by LA neighborhood.
The number of DV crimes mapped across the two years shows (green lines) a slight seasonal pattern to its movement over time. More DV occurrences occurred in the summer months versus between Oct – Mar of the year. Applying ARIMA models to DV trends (autocorrelation and partial correlation chart with 52 lags below) shows a relationship of time with DV occurrences, be it on the weak side. It would be helpful to parse and segment DV data from the city across neighborhoods and between weekends and weekdays. DV occurrences at the end of 2014 are higher than at the end of 2013. The mean of DV count crimes in 2013 (28) is lower than the 2014 mean (34). Then comparing the two years side by side of DV in 2013 and 2014, there are similar increases in Feb, May, July.
The increase in crime over summer months is common for criminal activity generally, probably the result of the heat’s effect on human behavior, school-aged children being out of school, and less preoccupations for families. Below in blue lines are other types of LA crimes over the two year period. Crimes with different root causes and motivations were included, such as battery, vandalism, burglary, theft identity, rape, and traffic related crimes. The year by year comparisons between different LA crimes shows similar difficulty with predicting and finding patterns. Lots more can be said about the non-DV crimes mapped below.
Time series trends can be stronger with different level of granularity, such as year, month, second, weekdays, or holidays. The weekends and weekdays time series are provided as a comparison and potentially simplifies the view of seasonality and trend analysis (pink graphs). Graphs show an increase of volume from 2013 to 2014. There are less weekday occurrences then weekend occurrences, however the outliers and spikes are higher on weekdays. More occurrences occur overall on the weekends.
Finally, I compared DV occurrences by neighborhoods in Los Angeles. There are lower occurrences on average in Rampart and West LA and more variation over time in 77th street and Pacific. There are inequities between neighborhoods but they are not as stark as I initially thought. For instance 77th street average occurrences is 42 per day compared with 30 occurrences per day in West LA. The difference of 12 is significant on one hand, 50% more occurrences happen in the 77th street neighborhood than West LA. On the other hand, an average cannot tell the whole story, since on any given day DV can occur more in West LA than 77th street.