Similarity Report ID: oid:3618:52481156

PAPER NAME

Paper(Imran).docx

WORD COUNT CHARACTER COUNT

2055 Words 11264 Characters

PAGE COUNT FILE SIZE

10 Pages 1.2MB

SUBMISSION DATE REPORT DATE

Feb 19, 2024 11:41 AM GMT+5:30 Feb 19, 2024 11:41 AM GMT+5:30

16% Overall Similarity

The combined total of all matches, including overlapping sources, for each database.
10% Internet database 06% Publications database

Crossref database Crossref Posted Content database

0% Submitted Works database

Summary
 CRIME DATA RATE PREDICTION USING HYBRID MACHINE LEARNING
ALGORITHMS
ABSTRACT
Crime is a pervasive concerning problem in modern civilization, with a multitude of crimes
being committed on a day-to-day, causing great unrest among the general populace. Preventing
crime is therefore an essential task. In recent times, the significance of artificial intelligence
has been observed in almost every field, including crime prediction. To achieve this, it is
2
essential to maintain a comprehensive data of past crimes, as this info can be utilized for future
orientation. Envisaging the likelihood of upcoming crimes can help law administration
agencies prevent them before they occur, providing valuable strategic evidence based on
factors as time and location. Though, accurately predicting crime is a challenging task, given
the increasing rate at which crimes are being committed. Therefore, the development of
effective crime prediction and analysis methods is crucial in detecting and reducing future
crimes. To this end, many researchers have led trials to predict crimes using support vector,
Logistics regression, Decision tree and XGboost machine learning techniques.

Keywords: Crime Rate Prediction, Accuracy, Machine Learning Algorithms

1. INTRODUCTION
2
Crime rate prediction is a field of study that aims to use machine learning methods to predict
and analyse the incidence of criminal activity in a particular area with specific time. The goal
is to recognize outlines and tendencies in crime data, use this data to develop stratagems and
interferences that can help avoid crime. Crime rate prediction typically involves the collection
2
and analysis of large amounts of data, including crime, demographic, and other relevant data.
2
Machine learning procedures are used to recognize outlines in the data, and to progress models
that can predict the likelihood of upcoming criminal movement. Few key applications of crime
rate calculation and scrutiny consist of the progress of prognostic policing approaches. the
recognition of high-risk areas for lawbreaking, and the assessment of the efficiency of crime
inhibition programs. By providing law implementation agencies and legislators with correct
2 2
and timely data about crime trends and patterns, crime rate prediction and analysis can help to
improve public safety and reduce crime rates. We utilized a crime dataset from Kaggle to train
2
our model. We opted to use the RF algorithm to analyse the patterns of crime and predict future
2
crime rates based on crime type. Additionally, we conducted an investigation of crime rates
across states over time, examining the time and kind of crime. Through these methods, our
study provides a comprehensive and structured overview of trends in crime rates and evaluates
existing methodologies for predicting crime rates.

2
An summary of several crime prediction approaches, including arithmetic models, machine
learning models, for spatial analysis, is provided in this survey. The preprocessing methods and
data sources used in crime prediction studies are also covered in the article. Predicting Crime
with Machine Learning S. Kumar et al, A Review of Crime Prediction Using Machine Learning
2
Techniques (2020) - This review essences mainly on using machine learning techniques to
2
predict crime. An overview of the different machine learning algorithms that are used to analyse
the different kinds of data used in crime prediction studies. A Survey of Crime Prediction and
Analysis Using Data Mining Techniques N. H. Mustapha et al.'s (2020) - This survey offers an
10
overview of data mining methods used for crime analysis and prediction. The paper covers
various data mining methods, including clustering, classification, and association rule mining,
that are applied in crime prediction studies.

A Systematic Review and Analysis of Benefits and Drawbacks B. K. Johnson et al., "Predictive
Policing: A Systematic Review and Future Directions" (2021) - This systematic review focuses
particularly on predictive policing, a contentious method of predicting crime that uses machine
learning algorithms to pinpoint neighbourhoods and people who are most likely to commit
crimes. In addition to outlining the present state of predictive policing study, the paper also
discusses some of its potential advantages and disadvantages. Exploring Spatial Analysis
Techniques in Crime Prediction. M. J. Beck and Y. B. Choi's 2017 article "Spatial Analysis and
Crime Prediction: A Review of the Literature" - The methods of spatial analysis employed in
studies of crime prediction are the particular focus of this review of the literature. The article
gives an summary of several spatial analysis procedures in addition to the kinds of data that
were used in these studies.

III. PROPOSED SYSTEM AND ARCHITECTURE

Crime hotspot estimate is a method used to recognize areas someplace upcoming crime actions
are likely to happen. In this study, the RF algorithm is applied to predict crime trials. The RF
algorithm has two aspects of randomness. The primary aspect is the use of bagging algorithm
9
to arbitrarily select training sample, the second part includes the assortment of a random split
5
attribute set. The best split mode off attributes is used to split the nodes. The multi decision tree
formed by the random forest algorithm ultimately determines the ending cataloguing outcome
 2
based on the tree classifier. The target of this research is to grow a model for foreseeing crime.
The training dataset is used to produce the model, and the test dataset is used to verify it. Based
2
on its accuracy, the algorithm used to teach the model is selected. The Random Forest algorithm
2
is applied in this situation to prediction crime. The dataset is pictured to analyse the crimes,
that have occurred in the nation, which can assist in finding trends and potential hotspots for
crime events in the future.

4
3.1 Benefits of Proposed System
The purpose of this research is to improve the earlier proposed estimate framework by using
4
alternate crime mapping and feature engineering approaches, and to deliver an open-source
execution that can be applied by police analysts for additional effective predictive policing.
This study aids law enforcement agencies in India by providing more accurate crime prediction
and detection, ultimately leading to a reduction in the overall crime rate.

3.2 Proposed System Architecture

Figure: 4.1: system architecture

IV. Module Description

The following phases are implemented for finding the results. Every phase is crucial for getting
accurate results.

• Collection of data
• Description of dataset
 • Pre-processing of data (loading the data)
• Selection of model
• Analysis of data
1
• Check Accuracy on test set
• Saving the Trained Model for further testing
4.1 Collection of data
3 3
Data collection is the primary substantial step to the genuine formation of a machine learning
3
model. Tata collection step is a vital phase influenced by more and better data; we can collect.
The data can be collected using a diversity of approaches, counting physical interferences, web
3
scraping. Machine Learning Algorithms for prediction Crime Hotspots from Kaggle and other
sources.
1
4.2 Dataset Description
There are distinct data points in the dataset. The dataset contains 15 columns, every is detailed
given below.
State: One of Indian state
District: A particular district
Year: 2001-2023
Murder: Murder rate
Rape: Rape rates
Theft: Theft rates
Total number of crimes: Whole crime rate

Figure 1: Crime Dataset

 V. RESULTS AND ANALYSIS
7
5.1 Random Forest regressor and Decision tree regressor
7
In our experiment we used random forest regressor and decision tree regressor combined to
give 95%+ accuracy of classification. The following figures 1 and 2 shows the login page and
theft crime rate.

Figure 1: Login page

Figure 2: Total Theft Crime Page

 Figure 3: Total Murder Crime Page
The figure 3 shows the murder crime rate of particular state.

Figure 4: Total Crimes in the state

The figure 4 shows the crime rate of Tamilnadan, Delhi, Kerala and Puducherry states.
 Figure 5: Total Crimes during the year 2017 to 2023
The above figure 5 shows the different types of crimes represented in the form of percentages.

5.2 Support Vector and XGBoost regressor

In our experiment we used support vector regressor and XGBoost regressor combined to give
different values of crime rates like theft, murder and rape etc. The following predictions of
SVR and XGB Regressor model building. Each crime predicts individually with help of
Support Vector regressor and XCBoost regressor.

The following individual predictions for rapes, murders, theft and over all crime rate. Here we
check the mean square error rate of support vector and XGBoost algorithms. In case 1 of rapes
prediction XGBoost algorithm generate minimum MSE. In case 2 murders SVM generate less
MSE. In case 3 theft XGBoost generate less MSE. In case 4 overall crime prediction XGBoost
generate less MSE.

Case 1: Handling rapes prediction

SVM Mean Squared Error: 85.26
XGBoost Mean Squared Error: 76.77
Case 2: Handling murder prediction
SVM Mean Squared Error: 53.23
XGBoost Mean Squared Error: 117.0
Case 3: Handling theft prediction
SVM Mean Squared Error: 190.4
XGBoost Mean Squared Error: 54.4
Case 4: Handling total crime prediction
SVM Mean Squared Error: 93.6
XGBoost Mean Squared Error: 81.6

5.2 Comparative Study

If you observe table 1, different machine learning algorithms used for calculating accuracy rate
with crime dataset. In this table shows ensemble or hybrid learning and individual classification
applied on crime dataset. Random Forest tree algorithm predict the higher accuracy 97.7% and
ensemble learning of RF and DT predicts the accuracy rate is 95.5%. Compare to all machine
learning algorithms, RF generate highest accuracy rate. In ensemble learning of our research
RF and DT predict the highest accuracy compare to SVM and XGBoost algorithms.

8
Table 1: Accuracy of different Machine Learning Algorithms
Algorithm Accuracy Rate
Random Forest 97.7%
Decision Tree 91.5%
Gradient Boosting 92%
Support Vector Machine & XGBoost 93%
Random Forest & Decision tree 95.5%
 VI. CONCLUSION
Preventing crime is therefore an essential task. In recent times, the significance of artificial
intelligence has been observed in almost every field, including crime prediction. To achieve
2
this, it is essential to maintain a comprehensive data of past crimes, as this info can be utilized
for future orientation. Envisaging the likelihood of upcoming crimes can help law
administration agencies prevent them before they occur, providing valuable strategic evidence
based on factors as time and location. Though, accurately predicting crime is a challenging
task, given the increasing rate at which crimes are being committed. Therefore, the
development of effective crime prediction and analysis methods is crucial in detecting and
reducing future crimes. To this end, many researchers have led trials to predict crimes using
support vector, Logistics regression, Decision tree and XGboost machine learning techniques.
 5
REFERENCES
[1] U. Thongsatapornwatana, ``A survey of datamining techniques in analysing crime patterns,''
1
Proc. 2nd Asian Conf. Defence Technol. (ACDT), Jan. 2016, pp. 123128.
[2] J. M. Caplan, L. W. Kennedy, and J. Miller, ``Risk terrain modelling: Brokering
criminological theory, GIS methods for crime forecasting,'' Justice Quart., vol. 28, no. 2, pp.
360381, Apr. 2011.
[3] M. Cahill and G. Mulligan, ``Using geographically weighted regression to explore local
crime patterns,'' Social Sci. Comput. Rev., vol. 25, no. 2, pp. 174193, May 2007.
[4] A. Alme hmadi, Z. Joudaki, and R. Jalali, ``Language usage on Twitter predicts crime rates,''
in Proc. 10th Int. Conf.Secur. Inf. Netw. (SIN), 2017, pp. 307310.
[5] H. Berestycki and J.-P. Nadal, ``Self-organised critical hot spots of criminal activity,'' Eur.
J. Appl. Math., vol. 21, nos. 45, pp. 371399, Oct. 2010.
[6] K. C. Baumgartner, S. Ferrari, and C. G. Salfati, ``Bayesian network modeling of offender
behavior for criminal proling,'' in Proc. 44th IEEE Conf. Decis. Control, Eur. Control Conf.
(CDC-ECC), Dec. 2005, pp. 27022709.
[7] W. Gorr and R. Harries, ``Introduction to crime forecasting,'' Int. J. Fore- casting, vol. 19,
no. 4, pp. 551555, Oct. 2003.
[8] W. H. Li, L.Wen, and Y. B. Chen, ``Application of improved GA-BP neural network model
in property crime prediction,'' Geomatics Inf. Sci. Wuhan Univ., vol. 42, no. 8, pp. 11101116,
2017.
[9] R. Haining, ``Mapping and analysing crime data: Lessons from research and practice,'' Int.
J. Geogr. Inf. Sci., vol. 16, no. 5, pp. 203507, 2002.
[10] S. Chainey, L. Tompson, and S. Uhlig, ``The utility of hotspot mapping for predicting
spatial patterns of crime,'' Secur. J., vol. 21, nos. 12, pp. 428, Feb. 2008.
[11] S. Chainey and J. Ratcliffe, ``GIS and crime mapping,'' Soc. Sci. Comput. Rev., vol. 25,
no. 2, pp. 279282, 2005.
[12] L. Lin,W. J. Liu, andW.W. Liao, ``Comparison of random forest algorithm and space-time
kernel density mapping for crime hotspot prediction,'' Prog. Geogr., vol. 37, no. 6, pp. 761771,
2018.
6
[13] Ch. Shravya, Pravallika and Subhani Shaik,” Heart disease prediction using Machine learning
Techniques”, International Journal of Innovative Technology and Exploring Engineering, Vol. 8,
Issue 6, 2019.
 Similarity Report ID: oid:3618:52481156

16% Overall Similarity

Top sources found in the following databases:
10% Internet database 06% Publications database

Crossref database Crossref Posted Content database

0% Submitted Works database

TOP SOURCES
The sources with the highest number of matches within the submission. Overlapping sources will not be
displayed.

ijsrset.com
1 3%
Internet

Rijul Kumar Srivastava, Anuj Gupta, Gaurav Sharma. "Forecasting Crim...

2 1%
Crossref

Vijayalakshmi G. V. Mahesh, Shilpa Hiremath, Chandra Prabha R.. "cha...

3 1%
Crossref

ijarcce.com
4 1%
Internet

Xu Zhang, Lin Liu, Luzi Xiao, Jiakai Ji. "Comparison of Machine Learnin...
5 <1%
Crossref

join.if.uinsgd.ac.id
6 <1%
Internet

Suraya Masrom, Thuraiya Mohd, Nur Syafiqah Jamil, Abdullah Sani Ab...
7 <1%
Crossref

dokumen.pub
8 <1%
Internet

Sources overview
 Similarity Report ID: oid:3618:52481156

jespublication.com
9 <1%
Internet

Judith K. Jones. "The role of data mining technology in the identificatio...

10 <1%
Crossref

Sources overview