Clustering of Indian States Based on Crime Incidences and Predicting Crimes Therein
Keywords:
crime analysis, crime in Indian states, crime prediction, hierarchical clustering, key predictors of crimes, , multiple linear regression, police data, socioeconomic variableAbstract
- To be pro-active in crime regulation, it is
crucial to monitor and analyse crime incidences. Socioeconomic and police data from 36 Indian states and
union territories are considered for the period 2014-
2018 to observe incidences of different types of crime.
Certain crime incidences have shown an alarming rate
of increase over this period, such as crime against
women and children (13% increase), crime against
senior citizens (30% increase), violent crimes (29%
increase) and dowry related crimes (31% increase).
Different states show different trends and intensities
on different crime types making analysis of crime
incidences more complex. Our study aims to cluster
states on several crime parameters and predict specific
crime incidences like murder, rape, dowry deaths,
cheating, etc. Hierarchical clustering using
agglomerative method is used to group states based on
features related to crime. Multiple regression with
feature selection technique is applied to identify key
predictors of crime, which includes number of schools,
number of jails, total and female police strength,
number of police stations, police budget and police
vehicles. Separate regression models are built for
different types of crime, with R-squared ranging from
0.60 to 0.76. While total police strength is a key
deterrent for murder cases, female police strength is a
key deterrent for dowry deaths, number of schools and
police vehicles are key deterrents for rape cases.
Setting up a robust education system, improving police
infrastructure and strengthening trained and mobile
police force will curb specific crimes like murder, rape,
dowry deaths etc., thereby making India safer and
more business friendly.