Crime is one of the biggest violations that has been not yet completely solved ever since the evolution of human race. In
order to solve this, crime analysis and prediction is one of the methods. Crime analysis is a scientific way of developing effective
strategies to prevent crime in future. In this project the crime analysis and prediction is done using different clustering approaches for
and various regression methods. DBSCAN and k-means clustering methods are used for analysis and regression methods such as ridge,
naïve Bayes and linear are used for prediction. Silhouette coefficient is used to determine the efficiency of the clustering methods. The
error values from the regression are determined using root mean square method. The crime data is extracted from State Crime Records
Bureau (SCRB) of Tamilnadu, India. It contains crime information about 38 different cities and districts. With the help of this
approach, crime can be predicated and reduced it in the future.
Raghavendhar T.V : Department of CSE, SRM Institute of Science and Technology,
Vadapalani, Chennai, Tamilnadu, India.
Joslin Joshy : Department of CSE, SRM Institute of Science and Technology,
Vadapalani, Chennai, Tamilnadu, India.
Mahaalakshmi R : Department of CSE, SRM Institute of Science and Technology,
Vadapalani, Chennai, Tamilnadu, India.
Ashutosh Soni M : Department of CSE, SRM Institute of Science and Technology,
Vadapalani, Chennai, Tamilnadu, India.
DBSCAN, k-means, linear regression, naive Bayes regression and ridge regression methods
DBSCAN, k-means, linear regression, naive Bayes regression and ridge regression methods clustering methods are implemented and their
performance is tested based on accuracy. On comparing
their performance the DBSCAN clustering has high
accuracy for the given dataset and forms effective clusters.
Table 2 shows Linear, Ridge and Naïve Bayes regressions
and their corresponding R- squared error value. On
comparing the different values and considering the
accuracy of our model, naïve Bayes regression show better
results as the values are closer to 1. Thus, this system will
help law enforcing agencies, police officials and general
public in enforcing laws and providing necessary
protection in areas that are vulnerable to crime.
[1] S.Sivaranjani, Dr.S.Sivakumari, Aasha.M. Crime
Prediction and Forecasting in Tamilnadu using
Clustering Approaches
[2] Nelson Baloian, Col. Enrique Bassaletti , Mario
Fernánde , Lt. Col. Oscar Figueroa, Pablo
Fuentes, Raúl, Manasevich, Marcos Orchard,
SergioPeñafiel1, joseA.Pino1, Mario Vergara
Crime Prediction using Patterns and Context.
[3] Xiafenwang, Mattew S. Gerber and Donald
E.Brown. Automatic Crime Prediction using
Events Extracted from Twitter Posts.
[4] Nurul Hazwan Mohammad shamsuddin, Nor
Azizah Ali, Razana Alwee. An Overview on
crime prediction Methods.
[5] Mohammad A, Tayebi, UweGlasser, Patrica L.
Brantingham. Learning where to inspect:
Location learning for crime prediction
[6] Coral Featherstone. The relevance of social media
as it applies in South Africa to crime prediction
[7] https://towardsdatascience.com/how-dbscanworks-
and-why-should-i-use-it-443b4a191c80
[8] https://www.datascience.com/blog/k-meansclustering
[9] http://www.stat.yale.edu/Courses/1997-
98/101/linreg.htm
[10] http://www.statsoft.com/Textbook/Naive-Bayes-
Classifier
[11] https://onlinecourses.science.psu.edu/stat857/node
/155
[12] https://study.com/academy/lesson/what-iscriminology-
definition-history-theories.html
[13] https://www.simplilearn.com/introduction-tomapreduce-
quick-tutorialvideo?
utm_source=google&utm_medium=cpc&ut
m_content=dsa&utm_term=&utm_campaign=sear
ch-dsa-generic-in-adgroup-allwebpages&
gclid=EAIaIQobChMI7bW05Iiv2gIVk
yQrCh1vgAwcEAAYASAAEgImg_D_BwE
[14] http://bigdata-madesimple.com/what-isclustering-
in-data-mining/
[15] http://www.comp.dit.ie/btierney/Oracle11gDoc/da
tamine.111/b28129/regress.htm
[16] http://www.tnpolice.gov.in
[17] G pavai T.V geetha new crossover operators usng
dominancea and co-dominance priciples for faster
convergance and genetic algorithm.