CRIME RATE PREDICTION AND ANALYSIS

USING MACHINE LEARNING

Anshul Kumar Ansh Bansal Ayaan Khan

Department if Computer Department if Computer Department if Computer
Science & Engineering Science & Engineering Science & Engineering
(Artificial Intelligence & (Artificial Intelligence & (Artificial Intelligence &
Machine Learning) Machine Learning) Machine Learning)
ABES Engineering College ABES Engineering College ABES Engineering College
Ghaziabad, India Ghaziabad, India Ghaziabad, India
Email:rajputanshul07082gmail Email:anshbansal288203@gm Email:ayaankhan958231@gm
.com ail.com ail.com

Dr.ROHIT
Associate Professor
ABES Engineering college
(Ghaziabad)

( ABSTRACT )

A crime is a deliberate act that can cause physical or data, criminal patterns can be identified, and
emotional distress, as well as property damage or predictions about future crimes can be made.
loss, and can lead to discipline by a state or other However, due to insufficient evidence, some crimes
authority based on the severity of the offense. The remain unreported and unsolved, making the task of
number and types of criminal activities are finding criminals challenging.
increasing at an alarming rate, prompting agencies Social media can be utilized to monitor criminal
to develop effective methods for preventative activity, as users often post about their surroundings
measures. Traditional crime-solving techniques are on these platforms.
now proving to be slow-paced and less effective in
the current scenarios.
To assist law enforcement, information can be ( INTRODUCTION )
extracted from text sources through social media
text analytics. Social media platforms, as text-based Crime rates are steadily increasing due to the
datasets, can significantly benefit from text advancements in modern technologies and
analysis. In today’s society, crime issues are sophisticated methods that enable criminals to
prevalent, and social media is no exception. Crime carry out illegal activities. According to the Crime
has negatively impacted economic growth and Record Bureau, incidents of burglary, arson,
quality of life. By examining and analyzing past murder, sexual abuse, gang rape, and other crimes
 are on the rise. Crime data is gathered from The primary goal of this project is to forecast crime
various blogs, news outlets, and websites to create rates and identify those that will occur in the
a comprehensive crime report database. The future. Authorities can use this information to take
preventative measures and lower crime rates. The
knowledge acquired from data mining techniques
project employs Multi Linear Regression to predict
helps in reducing crimes by identifying culprits
the relationship between crime types (independent
more efficiently. variable) and the year (dependent variable). To help
Security is a vital component of our lives, as our detectives solve crimes more quickly, the system
basic needs cannot be met unless we feel safe. Thus, will convert criminal information into a regression
having a sense of security is crucial for achieving problem. Crime analysis will identify trends based
our goals, whether they are shared or personal. on available data. Using multi-linear regression
algorithms, the frequency of crimes can be
Criminal activity is a significant social issue that
forecasted based on the territorial distribution and
negatively impacts society. Local authorities and crime recognition from the data.
residents are increasingly concerned with
3. Problem Statement
identifying crime hotspots and pinpointing recent
The aim of this project is to predict crime
crimes in specific areas. Additionally, people living occurrences using features present in the dataset,
in busy environments are constantly seeking ways which is extracted from official sources. By
to enhance safety and build trustworthy connections utilizing machine learning algorithms in Python,
with neighbors. the types of crimes likely to occur in a particular
area can be predicted.
Different crime acts include:
Domain Overview
• Act 379: Robbery
Machine Learning is a widely-used technique for
• Act 13: Gambling
predicting the future or classifying information to
• Act 279: Accident assist in decision-making. Machine learning
• Act 323: Violence algorithms learn from past experiences and
analyze historical data to identify patterns and
• Act 302: Murder
make future predictions. Data is the backbone of
• Act 363: Kidnapping machine learning algorithms. By training these
The output is the classification of the likely crime algorithms on historical data, more data can be
that has occurred. We experiment with various generated. For example, Generative Adversarial
classification algorithms, such as K-Nearest Networks (GANs) are an advanced concept in
Neighbors (KNN), Decision Trees, and Random machine learning that learns from historical
Forests. Multiple classification tasks are performed, images to generate new images. This concept is
also applied to speech and text synthesis.
including predicting which of the six classes of
Machine learning has thus opened vast potential
crimes are likely to occur and differentiating for data science applications.
between violent and non-violent crimes.

( EXISTING SYSTEM )

1. Scope of the Project Upon examining and understanding various

The current work focuses on two main objectives: distinct surveillance methods used by the police,
• Predicting crime surges and hotspots. the significance of each method becomes
• Understanding criminal behavior patterns to evident. Each surveillance method can perform
well independently and yield satisfactory results,
aid in criminal investigations.
albeit for a specific characteristic. For instance,
2. Objective of the Project a StingRay device is effective only when the
suspect is using a phone that is switched on.
Therefore, its usefulness is limited to situations 2. Preprocessing: Techniques are applied to
where the stakeout location information is accurate. clean the dataset.
Based on this understanding, technology continues 3. TF-IDF Calculation: TF-IDF values are
to evolve, producing smarter ways to conduct
surveillance. The introduction of deep learning, calculated for each pre-processed post.
machine learning, and computer vision techniques 4. Filtering and Classification: An
has offered new perspectives and innovative SVMbased filter is applied to remove
methods for surveillance activities. unrelated data, followed by using a random
forest classifier to categorize the data. The
main steps include formatting, cleaning, and
( MACHINE LEARNING )
sampling. The cleaning process involves
removing or fixing incomplete data.
Machine Learning is an interdisciplinary field that
combines computer science, mathematics, and
statistics. Statistics is crucial for making inferences (Machine learning)
from data, mathematics aids in developing machine
learning models, and computer science is used to 7 Steps of Machine Learning
implement algorithms. However, building models
alone is not sufficient; optimization and tuning of 1. Gathering Data: Collecting the necessary
the model are necessary to achieve accurate results. data for analysis.
This involves adjusting hyperparameters to reach 2. Preparing the Data: Cleaning and
the best possible outcome. As the world evolves, so
organizing the data for processing.
do the needs and requirements of people. We are
currently experiencing a fourth industrial 3. Choosing a Model: Selecting an
revolution driven by data. To extract meaningful appropriate machine learning model based
insights and understand how people and systems on the data and the problem.
interact with data, computational algorithms are 4. Training: Training the model on the dataset
essential. These algorithms process the data and to learn patterns.
provide beneficial results in various domains.
5. Evaluation: Assessing the model's
Machine Learning has transformed industries such
as medicine, healthcare, manufacturing, banking, performance using evaluation metrics.
and many others. 6. Hyperparameter Tuning: Adjusting the
Types of Machine Learning Machine model's hyperparameters to optimize
Learning algorithms can be classified into three performance.
types: 7. Prediction: Using the trained model to
• Supervised Learning make predictions on new data.
• Unsupervised Learning .
• Reinforcement Learning
> Introduction :
( PROPOSED SYSTEM )
In this blog, the workflow of a machine learning
A proposed approach utilizes SVM filtering project includes all the necessary steps to build a
methods to detect crime-related posts from social machine learning project from scratch. It also
media datasets. This process consists of four main covers data preprocessing, data cleaning, feature
phases: exploration, and feature engineering, highlighting
1. Extraction: Social media text posts related their impact on machine learning model
performance. Several pre-modeling steps can
to crimes are extracted.
help improve model performance.
 Python Libraries Required for the Task:
• Numpy (REFRENCES)
• Pandas
• Sci-kit Learn https://sist.sathyabama.ac.in/sist_naac/aqar_2022_
• Matplotlib 2023/documents/1.3.4/b.e-cse-batchno-119.pdf
Understanding the Machine Learning https://sist.sathyabama.ac.in/sist_naac/documents/

https:/www.sciencedirect.com/science/article/pii/S https://ijcrt.org/papers/IJCRT2009316.pdf
0952197618301520 https://www.ijert.org/research/crime-
detection-
Workflow: 1.3.4/1822-b.e-cse-batchno-242.pdf
The machine learning workflow can be defined by Science and Engineering
[1] Mugdha Sharma, Z-crime: A data mining
the following steps:
tool for the detection
1. Gathering Data: Collecting the necessary
ofsuspicious criminal activity based on decision
data for analysis. tree, IEEE, 2014,ISBN:978-
2. Data Preprocessing: Cleaning and 1-4799-4674-7/14
organizing the data for processing.
3. Researching Models: Identifying the best [2] Ubon Thansatapornwatana, A Survey of
Data
model for the type of data.
Mining Techniques forAnalyzing Crime
4. Training and Testing: Training the model on Patterns Second Asian
the dataset and testing its performance. Conference on
5. Evaluation: Assessing the model's DefenseTechnology ACDT, IEEE, 2016, ISBN:
performance using evaluation metrics. 978-1-5090-2258-8/16
What is a Machine Learning Model?
[3] Dr.Zakaria Suliman Zubi, Ayman Altaher
A machine learning model is essentially a piece of
Mahmmud, Using DataMining
code that becomes intelligent through training with
data. If poor-quality data is provided to the model, Techniques to Analyze Crime patterns in the
it will produce inaccurate or incorrect predictions, Libyan National Crime Data, Recent advances
often summarized as "garbage in, garbage out." in image, audio and signal processing.ISBN:
978-960-474-
350-6
( CONCLUSION )
[4]Shiju Sathyadeven, Deven M.S, Surya
Future enhancements of this research focus on Gangadharan. S, Crime Analysisand
training bots to predict crime-prone areas using prediction using data mining, IEEE, 2014
machine learning techniques. A method for
improving the detection of crime-related posts in
[5]Chung- Hsien Yu, Max W. Ward, Melissa
social media text messages is proposed. Initially, a
Morabito, Wei Ding
keyword-based filter is applied, followed by an
Crimeforecasting using data mining
SVM-based filter and random forest classification
to eliminate noise. According to current research, technique,11thIEEE
SVM achieves the highest accuracy among International Conference on
classifiers, making it the chosen method for this Electronics, Communication and Aerospace
study. Technology using-data-mining-
 IJERTV5IS010610 InternationalConference on ISBN: 978-1-4799-8553-1/15
Data mining workshop, IEEE, 2011,ISBN:978-0-
7695-4409-0/11 [9]Shyam Varan Nath, Crime Pattern Detection
[6]Fatih OZGUL, Claus ATZENBECK, Ahmet Using Data Mining,
CELIK, Zeki ERDEM,
International Conference on Web Intelligence and
Incorpating data sources and methologies for crime Intelligent Agent (WI-
data mining, IEEE,2011, IAT 2006 Workshops), IEEE, 2006, ISBN:0-7695-
ISBN: 978-1-61284-4577-0085-9/11 2749-3/06

[7]National Crime Records Bureau Website. [10]Jiawei Han, Micheline Kamber, Jian Pei,
[online] Data Mining: Concepts and Techniques 3rd ed.

[8]Arunima S. Kumar, Raju K. Gopal, Data Mining [11]http://www.cs.waikato.ac.nz/ml/weka/.online

Based CrimeInvestigation
System: Taxonomy and Relevance, IEEE, 2015, [12]https://www.r-project.org/about.html.[online]
ICECA 2017
978-1-5090-5686-6/17/$31.00 ©2017 IEEE 229 [13]http://r-statistics.co/Linear-
Regression.html.[online]