INTRODUCTION

 Crime has long been a significant challenge for societies

worldwide, impacting economic stability, public safety, and overall
quality of life (Tambi, 2019).
 As urbanization growth and populations increase, the complexity
of preventing and managing criminal activities has also increased.
 Traditional methods of crime analysis and prevention often rely on
manual investigation and reactive strategies, which can be time-
consuming and less effective in addressing dynamic criminal
patterns.
 With developments in technology, there is an crucial need to
connect data-driven approaches to enhance crime prediction and
prevention.
 In this research, Crime study, Prediction Using Machine Learning
and Deep Learning aims to address these challenges by
leveraging advanced computational techniques to analyze
historical crime data and predict potential future incidents
(Nowshin Tasnim, 2022).
 The integration of machine learning (ML) and deep learning (DL)
algorithms provides a robust framework for identifying crime
trends, detecting patterns, and forecasting high-risk areas
(Rajagopal RK, 2023).
Motivation of the
study
 As crimes commitment become increased from time to
time in our country and also the act of crimes increased,
it is necessary that, to develop an automated system
using deep learning techniques which classify and predict
the types of crime to reduce the acts of those crimes on
public or societies. Crime prediction is a complex problem
requiring advanced analytical tools to effectively address
the gaps in existing detection mechanisms. With the
increasing availability of crime data and through the
advancement of existing technology ,we have to provide
with a unique opportunity to study and investigates crime
detection using Deep learning methodologies. Apart from
predicting crime patterns, these algorithms can provide
valuable insights into crime trends and patterns. These
capabilities allow for deploying resources and tactics to
combat crime effectively.
 Previously, many scholars and researchers conduct a
research on crime prediction with various tools.
Some researcher uses statical analysis tools to
predict and classify crimes. The other researchers
uses only knowledge based system which uses a
recommendation system to forecast the rates of
crimes and make decision about that crime with
numerous limitation including, low accuracy result.
Few crime features and too low crime data set. As a
result,
 I am motivated to do this research to predict
and classify types of crime using deep learning
algorisms with better accuracy result, with more
crime features and also with more crime data set
than the previous researchers and scholars.
Statement of the problem
Crime has a significant impact on people's
lives, economic development, and overall
development. The trend indicates that the
crime rate is occasionally rising, and given
the volume of crime scenes, the coverage
of crime, and the complexity of the crime,
it is challenging to analyze effectively with
the resources currently available in police
stations (M. Saraiva, 2022). By making
predictions and adopting preventative
measures we can lower crime
 In previous years, many researchers have been
conducted a research on crime prevention using
statistical models to make an estimates in terms of
accuracy than human capacity to predict crime on
different location and time.
 However, the proposed previous researches and
scholars were not reliable, and that, the models they
were used should not be implemented with an
adequate accuracy.
 The previous study evaluates its performance on a
traditional hotspot policing method which was made
by the police and offer a clear fail of the processing
steps included to implement police predictive system
((Rahmatika, 2020). Crime is an increasing endemic
causes that contributes to social and economic
problems that in turn harmfully influence the growth
of a nation and its public societies (Matereke T,
2021).
Crime hotspot is highly increased and dispersed
rapidly in Nekemte city Administration , Which
includes, pick up societies mobile , money in
pocket, thefts of children, theft of one’s own
property and even killing the people to obtain
money and other goods (Endalew, 2017). Most of
the police station staffs in Nekemte town have
no enough information about the new pattern of
crime strategy including, where and when most
likely crime concentrated and committed (Chen,
2019). Crime prediction and identification are a
big problems that need to be addressed by the
police department, and given the vast amount of
crime data available (Greff, 2018).
 Deep learning techniques can assist to find
new patterns in various data sets and reveal
information that was previously undiscovered
(Borr, 2019). Various deep learning models
such as LSTM, CNN, RNN and DNN algorithms
are capable of predicting sequential data
(Matereke T, 2021) whereas LSTM is good for
categorical features (Borr, 2019). Combining
LSTM and RNN will be used to analyze both
nature of the attributes (Christian Janiesch,
2021). The writer (Chen, 2019) have
mentioned to improve the accuracy of the
model using complex Neural Networks and
Recurrent Neural Networks.
Research question
What types of crimes are
predicted by the ensemble
learning model?
What are the main ensemble
learning models selected for
crime prediction model
development?
To what extent ensemble learning
model better predict crime
patter?
Objectives of the study
 1General objective
 The general objective of this research is to design and
develop crime predictive model using deep learning
method.
 1.5.2 Specific Objectives
 To identify the possible crime types committed in
Nekemte city administration
 To compare and evaluate the performance of the
individual deep learning model.
 To conduct systematic literature review from secondary
source such as police station, experts and so on.
 To develop a crime predictive model using deep learning
techniques
 To test and measure the training capacity of developed
model on crime data set.
Scope of the study
The scope of this research mainly is about
designing and developing a crime predictive
model for crime analysis and prediction. We have
used a crime dataset of Nekemte city
administration from December 2011 up to
august 2017 of E.C . For the purpose of model
building and evaluating this data set will be
collected from cross sectional data of crime from
Nekemte police station and head quarter. On
other hand the scope of the research is based on
deep learning models using feed forward artificial
neural network, LSTM, RNN,CNN and the
combination of feed forward artificial neural
network and LSTM-RNN.
This proposed research work focused on
ensemble deep learning models by
combining the individual deep learning
models to add the capability of the model for
precisely predicting crime pattern and type
with efficient accuracy.
The other scope of the research is predicting
about six year crime data of Nekemte city
administration mainly committed in Nekemte
city. In our case, only the crime textual data
will used for crime analysis purpose, but not
include video and audio.
LITERATURE REVIEW AND RELATED WORK

This section represents the detailed

literature review work, which
involves the aim and focus
of the current research. Crime and
its types have diffed followed by
the process of crime
examination and analysis
approaches followed by process.
Lastly, we offer certain related
research works done by the
previous scholars and researchers.
Overview of crime prediction
 Typically,a law executions, citizens and
property protection, and control civil disorder
has
been entrusted to the police commissions that
is empowered by the state governments (Elluri,
2019). For such commission to be more
effective in its duties, Information is the vital
component
regarding catching the criminals and
investigating what did they do wrong (Hajela,
2020). And again, this information also helps to
in force law and impeachment. Through the
process the police force had the power to
legitimize the use of force.
 Specially law enforcement includes mostly on the policing
activities, which may comprise
emerging strategies to reply civil society request call for
help, identify suspects, search
for criminals who are at large, collect evidence on the
crime scene, investigate the crime
and finally up to make the case to the court (Jalil, 2020).
A suspect here means a person who is accused or
suspected that he/she committed the crime. Common
problems faced by police officers is predicting the future
crime so as to prevent the crime before it occurred.
Criminal
investigation and law enforcement bodies should be
enabled with technology to bring
solutions for their struggles and prevention activities of
organized crime by identifying and
targeting players of a crime done with complex network.
 The repetition of crime states that crime is a
societal phenomenon that fluctuates, and
these fluctuations have both spatial and
temporal dependencies (Dakalbab, 2020).
Accordingly, two types of features must be
taken into consideration in order to predict
future criminal activity in an area: temporal
characteristics, such as the times at which
criminal activity occurs and their
correlations, and spatial characteristics of
historical data, such as geospatial points of
manifestation of criminal activity and their
correlation.
In order to do this, several
strategies for preventing crime
have been implemented, with an
emphasis on intervention,
organized activities, and the
methods used to address a
nation's environmental, social,
and economic issues as well as
its legal system (Butt, 2020).
 Plan and schedule of
thework
No Activities
Dec. Jan. Feb. Mar. Apr. May. June.
1 Title selection
2 Title description
3 Literature review
Proposal writing
4 Proposal defense
5 Proposal approval
6 Data Collection
7 Data Entry
8 Analysis and Interpretation
9 Draft of theses preparation
10 Discussion and Conclusion
11 Recommendation and Future Works
12 Documentation of the Thesis
13 Submission of the Thesis
14 Thesis defense
 Budget plan of the research
study
NO Item Specifications Unit of quanti Unit Total
measureme ty cost cost In
nt ETB
Hardware Materials and resource
resource
needed to buy
1 External storage Portable storage disk drive One One 3500 3500
disk (HD) with HD storage
Storage capacity of one TB unit

2 memory D RAM (internal RAM) One module One 1500 1500

upgrading
internally to run
python
programming
3 Flash disk storage Flash Memory, One in One 800 800
Sony ,64GB ,USB 3.0 number
4 Transportation To Collect data from rural 1 month 30 150 4500
Agricultural land days

5 Communication Exchange of messages Voice 10 250 250

purpose card
6 Processor GeForce GTX 1080 X, 8Gb Video 1 2000 2000
graphics
6 Software Materials and resource
7 Python sw Paython 3.11 One pack 1 200 200
version DVD sw

8 Math lab VISUAL STUDIO - 1 500 500

9 MATLAB 2019 Full pack - 1 1500 1500

10 Development Classifiers Random 1 560 560

tools forest
Jupyter note book Machine 1 2000 2000
learning
Clustering like k-means Classifiers 1 200 200

Stationary purpose or
cost
11 Paper 4A paper 2 pack 2 400 800
12 Pen Red and black lexi 5 4 4 40 40
14 Print and copy Thesis document 200 page 2 400 400

15 Laminate and Thesis doc 4000 2 800 800

binding
16 Total budget 35000
reuired