Browse

Journals By Subject

Life Sciences, Agriculture & Food
Chemistry
Medicine & Health
Materials Science
Mathematics & Physics
Electrical & Computer Science
Earth, Energy & Environment
Architecture & Civil Engineering
Education
Economics & Management
Humanities & Social Sciences
Multidisciplinary

Books

Publish Conference Abstract Books
Upcoming Conference Abstract Books
Published Conference Abstract Books
Publish Your Books
Upcoming Books
Published Books

Proceedings

Events

Events
Five Types of Conference Publications

Join

Join as Editor-in-Chief
Join as Senior Editor
Join as Editorial Board Member
Join as Associate Editor
Become a Reviewer

Information

For Authors
For Reviewers
For Editors
For Conference Organizers
For Librarians
Article Processing Charges
Special Issue Guidelines
Editorial Process
Manuscript Transfers
Peer Review at SciencePG
Open Access
Copyright and License
Ethical Guidelines
Submit an Article
Research Article | | Peer-Reviewed

Evaluation of a Hyperparameter Tuned Random Forest Algorithm Based on Artificial Bee Colony for Improving Accuracy and Precision of Crime Prediction Model

Hauwa Abubakar*,  Souley Boukari,  Abdulsalam Ya’u Gital,  Fatima Umar Zambuk
Published in American Journal of Biological and Environmental Statistics (Volume 11, Issue 3)
Received: 6 June 2025     Accepted: 23 June 2025     Published: 11 September 2025
Views:       Downloads:
Download PDF
Abstract

Crime prediction is a crucial application of machine learning, enabling law enforcement to make proactive decisions. This research presents a novel crime prediction model that leverages a hybrid approach by tuning the hyperparameters of Random Forest (RF) classifier using the Artificial Bee Colony (ABC) optimization algorithm. The model was developed to improve the prediction accuracy and reliability of crime prediction systems by enhancing the performance of traditional machine learning classifiers. To validate the effectiveness of the proposed RF-ABC model, a comparative analysis against Decision Trees (DT), k-Nearest Neighbors (KNN), and the existing untuned Random Forest model was conducted. Experimental results demonstrate that the proposed RF-ABC model significantly outperforms the baseline models across multiple performance metrics. Specifically, the RF-ABC achieved an accuracy of 95%, precision of 90%, recall of 93%, and an F1-score of 90%. In comparison, the existing RF model yielded an accuracy of 81%, precision of 87%, recall of 84%, and an F1-score of 83%, while DT and KNN models recorded notably lower scores. DT obtained a PEI of 0.6900, PAI of 0.669 and RRI of 0.5200, while KNN has a PEI of 0.9647, PAI of 0.8670 and RRI of 0.5267, RF-ABC had the best result. PEI of 0.9800, PAI of 0.9000 and RRI of 0.7200. Crime prediction metrics show that These findings confirm that the integration of ABC with RF not only fine-tunes the hyperparameters efficiently but also enhances the model's predictive capabilities. The proposed hybrid approach shows promising potential for real-world crime analytic and decision support systems in law enforcement.

Published in American Journal of Biological and Environmental Statistics (Volume 11, Issue 3)
DOI 10.11648/j.ajbes.20251103.16
Page(s) 107-121
Creative Commons

This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited.

Copyright

Copyright © The Author(s), 2025. Published by Science Publishing Group
Previous article
Keywords

Crime Prediction, Hyperparameter, Artificial Bee Colony (ABC), Random Forest

1. Introduction
Crime prediction leverages criminology, data science and machine learning, to forecast criminal activities using historical data. While advancements in big data and Artificial Intelligence (AI) have enhanced predictive accuracy, challenges persist with dynamic crime data, including issues of over-fitting, under-fitting, and the lack of transparency in "black-box" models like deep learning
[1]
. This concept has been significantly bolstered by advancements in technology, particularly in the realms of big data and AI.
[2]
. In addition to preventing crime, accurate predictions can significantly improve the allocation of law enforcement resources
[3]
. Moreover, accurate crime prediction can aid in solving crimes more efficiently by identifying patterns and correlations that may not be immediately apparent to human analysts
[4]
. By uncovering these insights, law enforcement agencies can more quickly narrow down suspects, identify potential witnesses, and gather crucial evidence. This leads to faster resolutions of cases, increased clearance rates, and a greater sense of justice for victims.
Despite its potential benefits, crime prediction is not without its challenges and criticisms. One major concern is the risk of reinforcing existing biases in the data. If historical data reflects biased policing practices, predictive models may perpetuate these biases, leading to disproportionate targeting of certain communities.
Crime prediction process typically begins with data collection, where extensive datasets from various sources such as police reports, social media, economic indicators, and environmental sensors are gathered. This data is then cleaned, processed, and analyzed using sophisticated algorithms and statistical models.
[5]
By utilizing advanced data analytics and machine learning algorithms, such as Random Forests, law enforcement agencies can transition from reactive to proactive policing.
[6]
. However, achieving accuracy in crime prediction necessitates careful attention to data quality, algorithmic transparency, interpretability, and ethical considerations to ensure responsible and equitable use of these technologies.
The rapid evolution of technology and the increasing availability of data have catalyzed the development of predictive models across various domains, including crime prediction. However, despite numerous advancements, existing crime prediction algorithms often fail to meet the stringent requirements of accuracy, precision, and interpretability
[7]
. This inadequacy hampers their practical deployment and effectiveness in real-world scenarios where proactive crime prevention measures are crucial. Traditional crime prediction models frequently struggle with achieving high accuracy and precision
[8]
. This limitation is exacerbated by the dynamic and diverse nature of crime data, which can lead to either over-fitting or under-fitting. Consequently, these models may produce unreliable predictions, thereby reducing their utility for law enforcement agencies. While advanced machine learning techniques, such as deep learning, can achieve remarkable accuracy, they often function as "black boxes"
[9]
. The lack of transparency in their decision-making processes poses significant challenges for law enforcement agencies that require clear and interpretable insights to build trust, make informed decisions, and devise effective intervention strategies. Crime data is inherently complex and unpredictable, necessitating robust and adaptive modeling techniques. Current models often fall short in addressing this variability, resulting in sub-optimal predictions that hinder the development of proactive crime prevention strategies
[10]
.
The aim of the work is to develop an improved Random Forest algorithm that will enhance accuracy and precision of crime prediction model based on Artificial Bee Colony (ABC) hyper- parameter tuning and Recursive Feature Elimination with Cross Validation.
2. Related Works
The study provides an overview of machine learning methods relevant to crime prediction. It reviews state-of-the-art techniques that analyze crime data to identify patterns and develop predictive models, enhancing the accuracy and efficiency of crime forecasting. These techniques primarily analyze crime data to uncover patterns and generate predictive models.
[11]
investigates the influence of optimal hyper-parameters on machine learning algorithms for predicting heart disease, employing classifiers like Logistic Regression (LR), K-Nearest Neighbors (KNN), Support Vector Machines (SVM), Decision Trees (DT), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). The highest accuracy recorded was 87.9% with SVM, but the study faced challenges related to scalability and flexibility.
[12]
proposed hyper-parameter optimization for ensemble models in spam email detection using Grid-search for RF and XGBoost models. The results indicated high accuracy and sensitivity (97.78% and 98.09%, sensitivity of 98.44% and 98.84%, respectively), but it faced limitations in scalability and flexibility.
[13]
introduced a novel hyper-tuned ensemble Random Forest algorithm for detecting false basic safety messages in the Internet of Vehicles. The Random Forest (Ensemble RF) achieved better performance compared to KNN, DT, and other related works, but was constrained by computational costs due to exhaustive search methods.
[14]
, focused on crime analysis and prediction using machine learning approaches within the context of the Hossana Police Commission. The study employed RF, DT, and KNN. They indicated that RF outperformed the others but was prone to overfitting, data quality issues, and high computational demands.
[15]
, This research mapped crime risk terrains using machine learning, finding that RF outperformed KNN and kernel density estimation models at micro places but faced challenges related to interpretability and data quality.
[7]
, discussed how advancements in technology and data availability have impacted predictive modeling for crime prediction, emphasizing that many algorithms struggle to meet accuracy and interpretability standards.
[8]
, examined the challenges faced by traditional models leading to overfitting or under fitting issues.
[5]
compared KNN, DT and RF for crime prediction. The study analyzed various factors and patterns which the model identify areas or individuals that are at higher risk of being involved in criminal activities, Their work make conclusion that a boosted decision tree classifier performed better than DT, KNN and RF.
[16]
), proposed a method for improving the model robustness of flood hazard mapping based on hyper-parameter optimization of Random Forest using Grid Search, Bayesian Optimization, Random Search and Gauss process, the Bayesian Optimization was found to be more efficient than Grid and Random search, especially for high-dimensional hyper-parameter spaces. Table 1 Summarizes some of the relevant and related works.
Table 1. Summary of Related Works.

SN

Authors

Methods

Result obtained

Limitation

1.

[14]

RF, DT and KNN algorithms were used.

RF outperform DT and KNN having 81% Accuracy

Over fitting, interpretability, data quality, high computational requirements, and performed only classification,

2.

[15]

Random Forests, KNN and Kernel Density Estimation

RF greatly outperforms other crime prediction models. at micro places.

Interpretability, data quality, computational requirements, and ethical considerations

3.

[2]

XGBoost machine learning method and SHAP method

It explores the integration of environmental factors, such as crime opportunity theory, routine activity theory, rational choice theory, and crime pattern theory, into crime prediction models

The trade-off between Interpretability and Transparency

4.

[17]

Random Forest, Elastic Net, SVM

The study highlight the potential of ML models in identifying individuals at risk and devising proactive strategies to prevent criminal behavior among the population.

Refinement of risk prediction models is needed and variables with missing data were excluded from the analysis.

5.

[18]

Na¨ıve Bayes (NB), K-Nearest Neighbour (KNN), Support Vector Machine (SVM), and Random Forest (RF).

The study accurately predict risks assessment it leverage a spatial and temporal data, and adapt to new information, handle large datasets and provide decision support.

Limited model accuracy when patterns of crime change over time

6.

[5]

KNN, Decision tree, and Random forest

The study analyzed various factors and patterns which the model identify areas or individuals that are at higher risk of being involved in criminal activities.

Methods not Optimal
2.1. A Framework for Crime Prediction Using Random Forest with Parameter Optimization via Artificial Bee Colony Algorithm
This framework predicts crime occurrences by analyzing historical crime data, socioeconomic factors, and spatiotemporal trends using a data-driven approach. It involves preprocessing and cleaning data, clustering crime records with K-Means to improve feature relevance, and reducing dimensionality through auto-encoders to retain key patterns while lowering complexity. The data is split into training, testing, and validation sets, ensuring balanced crime categories. A Random Forest model is trained and optimized using the Artificial Bee Colony (ABC) algorithm to enhance prediction accuracy and efficiency. The optimized model is then compared with Decision Trees and k-Nearest Neighbors (kNN) based on accuracy, precision, recall, F1-score, and computational efficiency. Finally, the framework evaluates feature importance and the effectiveness of ABC optimization, providing insights for practical crime monitoring and prevention. This integrated approach combines advanced techniques to improve the accuracy and applicability of crime prediction models.
2.2. Simulation Framework for Crime Prediction Model Using Random Forest Parameters Optimized with Artificial Bee Colony Optimization
This framework details a virtual implementation of the crime prediction model, using structured steps for testing, validation, and analysis in a controlled environment. The process includes:
1. Data Preprocessing and Cleaning: Crime datasets are loaded, missing values imputed, outliers removed, numerical features normalized (Min-Max/Z-score), and categorical variables encoded (one-hot/label encoding).
2. Data Clustering with K-Means: Crime data is grouped by similarity, with optimal clusters determined via elbow/silhouette analysis.
3. Dimensionality Reduction via Autoencoder: An autoencoder neural network compresses data into lower dimensions, preserving essential patterns through reconstruction verification.
4. Data Splitting: Partitioning into training (70%), testing (20%), and validation (10%) sets, ensuring balanced crime category distribution.
5. Crime Prediction with ABC-Optimized Random Forest: A Random Forest classifier is trained, with hyperparameters (tree count, depth, etc.) optimized using the Artificial Bee Colony algorithm. Performance is evaluated on test/validation sets.
6. Model Comparison: Benchmarking against Decision Tree and KNN models using accuracy, precision, recall, F1-score, and computational efficiency.
7. Evaluation and Interpretation: Analyzing feature importance, ABC optimization impact, and deriving actionable insights for crime patterns. This end-to-end workflow leverages virtual tools and methodologies for robust algorithm testing, model validation, and hyperparameter optimization.
2.3. Research Gap
Existing crime prediction models including linear regression, decision trees (DT), and support vector machines (SVM) struggle with overfitting and poor interpretability, compromising their accuracy, reliability, and trustworthiness. These limitations hinder law enforcement and policymakers from making data-driven decisions. Although
[14]
improved accuracy by combining DT, Random Forest (RF), and K-Nearest Neighbors (KNN), critical issues persist:
1. Unoptimized RF hyperparameters exacerbate overfitting and interpretability challenges.
2. The opaque nature of RF models obscures prediction logic, limiting practical deployment.
To address these gaps, this research proposes a novel approach that enhances RF’s strengths while prioritizing accuracy and interpretability. The goal is to develop a trustworthy model for effective crime prevention and control.
3. Materials and Methods
Figure 1. Architecture of the proposed RF- ABC (Source:
[19]
).
Figure 2. Framework of the proposed RF- ABC (Source:
[19]
).
The proposed system improves upon existing models by utilizing historical crime data with features like crime type, location, time, demographics, and weather. Data preprocessing involved handling missing values, encoding categorical variables, and normalizing numerical features. An autoencoder reduced data dimensionality and extracted meaningful features, followed by K-means clustering to group similar crimes. The dataset was split into training (70%), testing (15%), and validation (15%) sets. Random Forest (RF) was trained using bootstrap sampling and feature randomness at each tree split to reduce overfitting. The Artificial Bee Colony (ABC) algorithm optimized RF hyperparameters. Fully grown decision trees captured complex patterns, with classification based on majority voting and regression using averaged predictions. The model was evaluated using the testing data. The performance criteria include accuracy, precision, recall and F1 score, PAI, PEI and RRI etc. Figure 1 shows the Architecture of the proposed RF- ABC model. Figure 2 shows the framework of the proposed RF – ABC Model.
3.1. Model Selection and Training
Random Forest is an ensemble learning method used for classification and regression. It builds multiple decision trees from bootstrapped samples of the data and combines their results to improve accuracy and reduce overfitting. Randomness is introduced by selecting a random subset of features at each split, which decreases correlation among trees. For classification (e.g., crime type prediction), the final decision is made by majority voting, while for regression (e.g., crime rate prediction), it averages the outputs of all trees. Model performance depends on hyperparameters like the number of trees, tree depth, and features considered for splits. The Artificial Bee Colony (ABC) algorithm is used to tune these hyperparameters effectively. The steps Random Forest and ABC algorithms are presented below:
1. Algorithm 1 RF
1) Initialize Random Forest Parameters (T, L, S, G)
2) For each tree draw a bootstrap sample with replacement from original dataset Dt (Some data points may be repeated, and some may be left out: out-of-bag samples)
3) Build and Train Decision Trees ht
At each node select random subset of m features based on chosen criterion
Split nodes using best feature
Continue splitting until stopping criteria is met
4) Aggregate Prediction from all trees
P = mode (ht, ht, ht, ht, ht, ht ………)
5) Output the final class or regression (Y)
2. Algorithm 2 ABC
1) Initialization: Initialize a population of potential solutions (bee colony). Xi= (1=1, 2, 3, n) and initialize ABC parameters Where each bee Xi = X0 + rand (0,1) * (Xmax – Xmin)
2) Determined the fitness of each search agent (Employed Bees Phase).
3) Each employed bee explores a new solution in its neighborhood V
V = Xi + α * (Xi – Xk) where Xk is a random solution and α is between [-1.1]
4) Onlooker Bees Phase perform step 3. Using a probability based on fitness.
5) Scout Bees perform step 2&3 if solution is not improved
6) While number of explorations are repeated (Iteration t < max_iterations):
7) For each search agent;
8) Modify the current search agent’s position.
9) Evaluate the fitness of all search agents
10) Modify Xo, Xb, and Xc
11) + + t
12) End while
3. Algorithm 3 Hybrid RF-ABC
The algorithm for the hybrid Random Forest Algorithm with hyperparameters optimized using artificial bee colony is shown in the steps below. The flow chart of the proposed model is shown in Figure 3.
Steps:
1) Initialization: Initialize ABC parameters and generate potential solutions (bee colony). Xi= (1=1, 2, 3, n) consisting of RF parameter (n_estimators max_depth, min_samples_split, min_samples_leaf) Where each bee Xi = X0 + rand (0,1) * (Xmax – Xmin)
2) Determined the fitness of each search agent (Employed Bees Phase)
(a) Initialize Random Forest Parameters (T, L, S, G)
(b) For each tree draw a bootstrap sample with replacement from original dataset Dt (Some data points may be repeated, and some may be left out: out-of-bag samples).
(c) Build and Train Decision Trees ht
(d) At each node select random subset of m features based on chosen criterion
(e) Split nodes using best feature
(f) Continue splitting until stopping criteria is met
(g) Aggregate Prediction from all tress P = mode (ht, ht, ht, ht, ht, ht ….)
(h) Output the final class or regression (Y)
3) Each employed bee explores a new solution in its neighborhood V
4) V = Xi + α * (Xi – Xk) where Xk is a random solution and α is between [-1.1]
5) Onlooker Bees Phase perform step 3. Using a probability based on fitness.
6) Scout Bees perform step 2&3 if solution is not improved
7) While number of explorations are repeated (Iteration t < max_iterations):
8) For each search agent;
9) Modify the current search agent’s position.
10) Evaluate the fitness of all search agents
11) Modify Xo, Xb, and Xc
12) + + t
13) End while
Figure 3. Flow Chart of the Proposed Model (Source:
[19]
).
3.2. Dataset
The Chicago Crime Dataset is a publicly available dataset that contains detailed records of crimes reported in the city of Chicago, Illinois, USA. It is maintained by the Chicago Police Department and is part of the city's open data initiative. The dataset is widely used for research, analysis, and predictive modelling in criminology, urban planning, and data science (19). The original dataset comprised 228004 records and 22 attributes. These attributes include the offender's name, ID number, sex, age, education status, job, and marital status, as well as the victim's sex, age, and the place and type of crime. The dataset includes incidents from 2010 through May 2023. The system collects relevant data sources, such as crime incident reports, socio-demographic data, and environmental factors. The collected data was then prepossessed, cleaned, normalized, and integrated to create a comprehensive dataset for analysis. Data was then divided in to training data and testing data. Table 2 shows the overview of the database and Table 3 presents the description of each feature in the dataset. The table has 12 categorical features and 10 numerical features.
Table 2. Overview of Database.

Item

Description

Name

Chicago crime

Data Source

Chicago crime dataset

Total Records

239, 559

Total Columns

24

Total Features

23

Class

5

Categorical Values

5

Numerical Values

19

Class Distribution

6

Missing Values

6,222

Year

16-94 years

Period

12 Months
Table 3. Description of Features from Database.

S/n

Feature

Type

Full Name

Units

1

ID

Number

Identification Number

Number

2

Case Num

Categorical

Number of Case

Number

3

Date

Categorical

Date crime was committed

Date

4

Block

Categorical

Block crime was committed

Block

5

fUCR

Number

classification system

Classification

6

Prim Type

Categorical

Primary Type of crime

Type

7

Description

Categorical

Description of Crime

Description

8

Location

Categorical

Location crime was committed

Location

9

Gender

Categorical

Sex offender

Male/Female

10

Arrest

Categorical

Arrest was made

True/False

11

Domestic

Categorical

Domestic crime or not

True/False

12

Beat

Number

Geographic area

Location

13

District

Number

District crime was committed

District

14

Ward

Number

Ward crime was committed

Ward

15

Community

Number

community

Community

16

Crime Code

Categorical

code of crime

code

17

X Cord

Number

x coordinates

Degrees

18

Y Cord

Number

y coordinates

Degrees

19

Year

Categorical

Year of crime

2022-2022

20

Updated

Categorical

updated on database

Yes/No

21

Latitude

Number

Longitude

Degrees

22

Longitude

Number

Latitude

Degrees
3.2.1. Data Preprocessing
The data used for model development was thoroughly cleaned and processed to ensure high quality, including handling missing values, removing noise and outliers, and converting file formats for compatibility. After cleaning, 228,004 records with 10 attributes were retained for the model development.
3.2.2. Feature Selection
To improve crime prediction accuracy, the system used Recursive Feature Elimination (RFE). Recursive feature elimination is a popular feature selection algorithm that is often used
[20]
. This process allows feature elimination to identify the most important features in the dataset and rank them in order of importance. Another advantage is that it is less sensitive to the choice of tuning parameter, which can lead to more stable and reliable results
[21]
.
3.3. Performance Evaluation Techniques
The proposed model integrates Artificial Bee Colony (ABC) optimization with Random Forest (RF) to improve crime prediction accuracy and precision. Its performance is evaluated using several key techniques.
3.3.1. Evaluation Metrics
The developed ensemble learning model’s performance was evaluated using accuracy, precision, recall, and F1-score, assessing its effectiveness in predicting crime rates and types. Compared to baseline and existing methods, integrating the Artificial Bee Colony (ABC) algorithm for hyperparameter tuning strengthened the Random Forest model, resulting in improved performance and prediction quality.
1. Accuracy Rate
Accuracy is a metric that measures how often a machine learning model correctly predicts the outcome. Accuracy is obtained by dividing the number of correct predictions by the total number of predictions
Accuracy Rate =TP+TNTP+FP+TN+FN(1)
2. Precision
Precision measures the accuracy of a model's positive predictions by calculating the ratio of true positives to all predicted positives. It is especially useful for imbalanced datasets, reflecting how well the model identifies the target class. It is given as;
Precision =TPTP+FP(2)
3. Recall
Recall is a metric that measures how often a machine learning model correctly identifies positive instances (true positives) from all the actual positive samples in the dataset. It is calculated by dividing the number of true positives by the number of positive instances. The latter includes true positives (successfully identified cases) and false negative results (missed cases).
Recall =TPTP+FN(3)
Recall is measured on a scale of 0 to 1 or as a percentage. The higher the recall, the better. A perfect recall of 1.0 is achieved when the model can find all instances of the target class in the dataset.
4. Predictive accuracy index (PAI)
The Prediction Accuracy Index (PAI) measures how accurately a model predicts crime hotspots by comparing the variance of estimated responses during review and development phases. Unlike other metrics, PAI accounts for the size of study areas, making it a widely used and effective measure for evaluating the spatial accuracy of crime predictions, reflecting both correct and incorrect hotspot identifications.
The PAI is calculated as follows:
PAI=TP+TNTP+TN+FP+FN(4)
Where:
1) TP: True Positives (correctly predicted crime hotspots)
2) TN: True Negatives (correctly predicted non-crime hotspots)
3) FP: False Positives (incorrectly predicted crime hotspots)
4) FN: False Negatives (incorrectly predicted non-crime hotspots)
Interpretation of PAI:
PAI > 0.5: Indicates good prediction accuracy, suggesting the model effectively identified both actual crime hot-spots and non-hotpots.
PAI = 0.5: Represents chance performance, suggesting the model does not significantly improve upon random guessing.
PAI < 0.5: Indicates poor prediction accuracy, suggesting the model frequently misidentified crime hot-spots.
5. Predictive efficiency index (PEI)
The PEI measures how well a forecasting algorithm does compare to how well it could have done in forecasting the number of crimes
[10]
. The PEI can be expressed as a proportion (varies from 0 to 1) or a percentage (varies from 0 to 100).
The Predictive Efficiency Index (PEI) evaluates crime forecasting models by measuring both the accuracy of predictions and the resources needed to produce them.
The PEI is calculated as follows:
PEI=TP+TNTP+TN+FP+FN*TC(5)
Where:
1) TP is the number of true positives (correctly predicted crimes)
2) TN is the number of true negatives (correctly predicted non-crimes)
3) FP is the number of false positives (incorrectly predicted crimes)
4) FN is the number of false negatives (incorrectly predicted non-crimes)
5) T is the total number of predicted crimes
6) C is the total number of actual crimes
The PEI ranges from 0 to 1, with a higher score indicating a better-performing model. PEI is a good measure because it considers both accuracy and resource efficiency. It is relatively easy to calculate and can be used to compare different models. On the contrast, PEI is sensitive to the distribution of crimes and PEI may not be suitable for all types of crime forecasting models. PEI also does not consider the severity of crimes.
6. Recapture Rate Index (RRI)
The Recapture Rate Index (RRI) is a metric used to assess the precision of crime forecasting models, focusing specifically on their ability to recapture crime hotspots in a future period. It considers changes in crime density over time, making it a valuable tool for evaluating long-term forecasting performance. RRI is a measure of the percentage of actual crimes that are correctly predicted. The RRI is calculated as follows:
RRI=HRR*HDHRR+HNR(6)
Where:
HRR: Hotspots Recaptured Ratio = Number of predicted crime hotspots that were actual hotspots
in the future period / Total number of predicted crime hotspots.
HD: Historical Density = Average crime density across the entire study area in the historical period.
HNR: Hotspots Not Recaptured Ratio = Number of predicted crime hotspots that were not actual hotspots in the future period / Total number of predicted crime hotspots.
Interpretation of RRI:
RRI > 1: Indicates an increase in crime density in predicted hotspots, suggesting the model accurately identified potential hotspots.
RRI =1: Indicates that no change in crime density in predicted hotspots, meaning the model did not predict future crime patterns effectively.
RRI < 1: Indicates a decrease in crime density in predicted hotspots, suggesting the model overestimated future crime activity.
The Recapture Rate Index (RRI) emphasizes precision, aiding resource allocation and long-term crime prevention by accounting for changes in crime density over time. It is simple to calculate and interpret but is sensitive to how crime hotspots and thresholds are defined. RRI is less suitable for short-term forecasts due to reporting delays and does not directly measure the accuracy of predicted crime counts.
[22]
.
3.3.2. Experimental Setup
The model was implemented in MATHLAB 2019 The model was then experimented using the dataset specified in the design. The system used is a laptop, laptop System Core (TM) i5-5200U, 2.20 GHz CPU, 8.00 GB RAM, 500 GB hard drive. ABC was used to obtain the optimal parameters of a RF model as a result enhancing its performance in crime prediction tasks. The Table 4 below shows the parameters for the RF (default) and the search range for ABC (minimum, and maximum values). Table 5 shows the ABC parameters.
Table 4. Default and the optimal parameters for the Proposed Model.

RF Parameters

Default

Minimum

Maximum

NumTrees

50

10

100

MaxNumSplits

2

1

100

MinLeafSize

1

1

5

MinParentSize

10

5

20

Max Iteration

100

10

1000
Table 5. ABC Parameters.

RF Parameters

Default

Maximum Iteration

100

Population of Bees

50

Onlooker Bees

50

Abandonment Limit

0.6

Acceleration Coefficient

1
The default parameter settings for the Random Forest are NumTrees 50, MaxNumSplits 2, MinLeafSize 1 and MinParentSize 10. The ABC searched for the optimal parameters setting between the minimal and maximum range for the parameters.
4. Results
The results are analyzed in two key phases: the performance of the standalone Random Forest model and the improvements achieved by incorporating the ABC optimization. The discussion includes a detailed evaluation of critical performance metrics such as accuracy, precision, recall, and F1-score.
4.1. Optimized RF Parameters Using ABC
The Artificial Bee Colony (ABC) algorithm optimizes Random Forest (RF) hyperparameters like the number of trees, tree depth, and features per split by mimicking honeybee foraging behavior to efficiently explore solutions. This fine-tuning improves model generalization, reduces overfitting, and enhances accuracy and precision, especially in complex tasks like crime prediction. ABC iteratively evaluates and refines candidate solutions, resulting in a more robust RF model that outperforms default or manually tuned versions.
Table 6. The default Random Forest Parameters and the Optimal parameters obtained after optimizing with ABC.

RF Parameters

Existing

Optimal (Proposed RF-ABC)

NumTrees

50

70

MaxNumSplits

2

4

MinLeafSize

1

5

MinParentSize

10

10
Here is a summary of the optimized Random Forest hyperparameters:
1) NumTree (Number of Trees): Increasing the number of trees from the default 50 to 70 improves model stability and performance, though gains diminish beyond this point and computational costs increase.
2) MaxNumSplits (Maximum Number of Splits): Allowing more splits helps the model capture complex patterns, but too many splits risk overfitting. The optimal value found is 2.
3) MinLeafSize (Minimum Leaf Size): Smaller leaf sizes capture finer details but may cause overfitting, while larger leaf sizes simplify the model and can improve generalization.
4) MinParentSize (Minimum Parent Size): it generally controls the minimum number of samples required to split a node, balancing model complexity and overfitting.
This minimum parent size controls the minimum number of observations required in a node for it to be split. Reducing this value leads to more complex trees; while increasing it can simplify the model. A value between 5 and 10 is generally effective. After fine tuning the ABC arrived at a value of 10.
4.2. Model Performance
Table 7 presents the performance of models developed, it compares the performance metrics of a standard DT, KNN and Random Forest (RF) models with RF model enhanced with the ABC optimization algorithm, the accuracy for the models DT, KNN, RF and RF-ABC are 74, 79, 81 and 95 respectively, while the precision for DT, KNN, RF and RF-ABC are 86, 81, 87 and 90 respectively. the recall obtained by the models Recall are 68, 70, 84 and 93 respectively. while the F1 score is 78, 69, 83 and 90 respectively. The Table 7 shows the results obtained. Figure 4 presents a bar chart showing the performance comparison of the four metrics for the crime prediction Models developed and tested.
Table 7. Performance Comparison the Proposed model with other models.

DT

KNN

Existing (RF)

Proposed (RF-ABC)

Accuracy

74

79

81

95

Precision

86

81

87

90

Recall

68

70

84

93

F1-Score

78

69

83

90
Figure 4. Performance comparison of the crime prediction models.
The RF-ABC model achieves an accuracy of 95%, which is significantly higher than the RF model's accuracy of 81%. This indicates that ABC optimization effectively improves the model's ability to make correct predictions across the dataset. The RF-ABC model performed better than RF, DT and KNN Models. Precision is slightly higher for the RF-ABC model (90%) compared to the standard RF (87%) and much better than DT (86%) and KNN (81%). This improvement reflects a reduced number of false positive predictions, suggesting that RF-ABC optimization helps in identifying the relevant features and reducing noise in the data. The RF-ABC model achieved the highest recall at 93%, outperforming the standard Random Forest (84%), Decision Tree (68%), and KNN (70%), indicating superior ability to identify true positives a crucial advantage in scenarios where missing actual cases has serious consequences. This strong recall is especially valuable for imbalanced datasets.
In terms of F1-Score, which balances precision and recall, RF-ABC again led with 90%, followed by RF at 83%, DT at 78%, and KNN at 69%. KNN’s low F1-Score suggests poor handling of noisy or complex data, while DT’s moderate performance may be due to overfitting. The Random Forest model improved generalization and reduced overfitting, but the addition of ABC hyperparameter tuning further enhanced both accuracy and precision, making RF-ABC the best overall performer.
In summary:
1. KNN (69%): Weakest, likely affected by noise or complexity.
2. DT (78%): Better, but prone to overfitting.
3. RF (83%): Stronger, with improved balance and generalization.
4. RF-ABC (90%): Best performance due to optimized hyperparameters, achieving the highest accuracy and precision.
4.3. Crime Model Performance
The Table 8 presents the performance of four crime analysis models DT, K-Nearest KNN, RF, and RF-ABC evaluated using three key crime-related indices: PEI, PAI, and RRI. Each index provides insight into how well these models predict, analyze, and assess crime trends.
Table 8. Performance Comparison Indices for the crime prediction models.

Parameter

DT

KNN

Existing (RF)

Proposed (RF -ABC)

PEI

0.69

0.7117

0.9647

0.98

PAI

0.6617

0.6832

0.867

0.9

RRI

0.52

0.5023

0.5267

0.72
Figure 5. Performance comparison of the performance indices of the crime prediction models.
4.3.1. Predictive Efficiency Index (PEI)
The Predictive Efficiency Index (PEI) evaluates how effectively a model distinguishes between crime and non-crime events. The RF-ABC model achieved the highest PEI (0.9811), demonstrating superior efficiency and accuracy in crime classification, thanks to ABC optimization fine-tuning Random Forest parameters. Standard Random Forest also performed well (PEI = 0.964) but was slightly less efficient without ABC optimization. KNN (0.7117) and Decision Tree (0.6900) had lower PEI scores, with KNN struggling on complex or overlapping data and DT being prone to overfitting, making both less reliable for accurate crime prediction. Overall, RF-ABC proved to be the most effective model for real-world crime prediction.
4.3.2. Predictive Accuracy Index (PAI)
PAI measures how well models predict police actions based on crime trends. The RF-ABC model achieved the highest PAI (0.9006) due to enhanced feature selection and bias reduction from ABC optimization, making it highly effective for crime prevention and patrol planning. Standard Random Forest also performed well (0.8670) but was slightly less accurate without hyperparameter tuning. KNN (0.6832) and Decision Tree (0.6617) showed lower performance, struggling with dynamic crime patterns and nonlinear relationships, respectively. Overall, RF-ABC is the most accurate and reliable model for forecasting law enforcement activities.
4.3.3. Recapture Rate Index (RRI)
The Recapture Rate Index (RRI) evaluates how well a model identifies and distinguishes crime risks across locations, times, and demographic groups. The RF-ABC model achieved the highest RRI (0.7267), showing superior ability to detect subtle crime patterns and variations, making it valuable for targeted crime prevention and resource allocation. Standard Random Forest performed moderately (0.5267) but was less effective at differentiating crime risks. Decision Tree (0.5200) and KNN (0.5023) had the lowest RRI scores, struggling to accurately capture crime risk variations, which limits their usefulness for strategic crime prevention planning.
4.3.4. Computational Resources
1. Training Time:
The training time for RF-ABC is 1291 seconds, compared to just 16 seconds for the standard RF model. This significant increase is due to the additional computational effort required by the ABC optimization algorithm, which searches for optimal hyperparameters or feature subsets. Table 9 shows the computational resources in terms of time and memory used by the developed models. The comparison of run time for various models is also presented in Figure 6.
Table 9. Comparison of models computational Resources.

DT

KNN

Existing (RF)

Proposed (RF-ABC)

Time (s)

4

12

16

1291

Memory (kb)

1.3

1.6

2.81

2.85
Figure 6. Graphical representation of Run Time of the crime prediction Models.
2. Memory Usage
From the Table 9 and Figure 7, it can be seen that RF-ABC models use more memory due to the iterative nature of the ABC optimization process and the storage of intermediate results during hyperparameter tuning. The RF-ABC model shows clear advantages in terms of predictive performance (accuracy, precision, recall, and F1-Score). This makes it suitable for applications where accuracy is paramount and computational resources are not a constraint. However, the higher training time could be a drawback in real-time or resource-constrained environments. Additionally, the memory requirements (if higher) may limit the applicability in devices with limited hardware.
Figure 7. Graphical representation of Memory Used for the Models.
3. Confusion Matrix for RF and RF-ABC Models
The confusion matrices for the Random Forest (RF) model and the RF model enhanced with ABC optimization (RF-ABC) show their crime prediction performance. The RF model predicted Burglary with 81.5% accuracy, Arson at 79%, and both Sex and Drug crimes at 80%, achieving an overall accuracy of 81.5%. In contrast, the RF-ABC model significantly improved predictions, achieving 95% accuracy for Burglary, 93.5% for Arson, 95.4% for Sex crimes, and 95.9% for Drug crimes, with an overall accuracy of 95.5%.
Figure 8. Confusion Matrix for predicted types using the RF-ABC crime prediction model.
5. Discussions
Ensemble Superiority: Both PEI and PAI metrics clearly show that ensemble methods (RF and RF-ABC) outperform the simpler DT and KNN approaches. The ensemble approach’s ability to aggregate multiple decision trees provides a more robust and accurate prediction model for crime.
1. Impact of Optimization: The RF-ABC model consistently outperforms the standard RF across all three metrics, most notably in risk reduction (RRI). This demonstrates the potential benefits of applying an optimization algorithm, such as the Artificial Bee Colony algorithm, to fine-tune the model parameters or feature selection process. The improvements in PEI and PAI also suggest that the optimization not only enhances efficiency and accuracy but also contributes to more reliable predictions in risk-sensitive applications.
2. Risk Reduction Challenges: While the RF-ABC model significantly improves the RRI, the relatively lower values of RRI for DT, KNN, and even the standard RF indicate that risk reduction remains a challenging aspect of crime prediction. This might necessitate further research and development, possibly incorporating additional risk management strategies. or more sophisticated modeling techniques, to better address uncertainty and minimize prediction errors in high-stakes environments.
6. Conclusion
The study demonstrates that for crime prediction tasks, the combination of ensemble learning with an optimization strategy RF-ABC enhances model performance across multiple evaluation metrics. These findings underscore the importance of not only selecting robust machine learning models but also optimizing them to better handle the complexities and risks associated with predicting criminal activities. The integration of ABC optimization significantly enhances the performance metrics of the Random Forest model, at the cost of increased computational time. For high-stakes applications where predictive accuracy is critical, the RF-ABC model is highly recommended. However, for less demanding tasks or real-time constraints, the standard RF model remains a viable option. Future work could explore additional optimization techniques or hybrid models to further refine these performance metrics, particularly the RRI, to support more effective and reliable crime prevention strategies.
Abbreviations

ABC

Artificial Bee Colony

ARFF

Attribute Relationship File Format

CSV

Command Separated Values

DT

Decision Tree

KNN

K - Nearest Neighbor

LR

Linear Regression

PAI

Predictive Accuracy Index

PEI

Predictive Efficiency Index

RRI

Recapture Rate Index

RF

Random Forest

SVM

Support Vector Machine
Author Contributions
Hauwa Abubakar: Conceptualization, Investigation, Methodology, Resources, Software, Writing – original draft
Souley Boukari: Project administration, Supervision, Validation, Writing – review & editing
Abdulsalam Ya’u Gital: Project administration, Supervision, Validation, Writing – review & editing
Fatima Umar Zambuk: Supervision, Validation
Funding
The research did not receive any specific grant from funding agencies in the public, commercial or non-profitable organizations.
Data Availability Statement
The data is available from the corresponding author upon reasonable request.
Conflicts of Interest
The authors declare no conflicts of interest.
References
[1] Adeyemi, R. A., Mayaki, J., Zewotir, T. T., & Ramroop, S. (2021). Demography and Crime: A Spatial analysis of geographical patterns and risk factors of Crimes in Nigeria. Spatial Statistics, 41, 100485.
[2] Zhang, H., Gao, Y., Yao, D., & Zhang, J. (2023). Interaction of Crime Risk across Crime Types in Hotspot Areas. ISPRS International Journal of Geo-Information, 12(4), 176.
[3] Bischl, B., Binder, M., Lang, M., Pielok, T., Richter, J., Coors, S., & Lindauer, M. (2023). Hyperparameter optimization: Foundations, algorithms, best practices, and open challenges. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 13(2), e1484.
[4] Zhao, C., & Cao, H. (2022). Risk Evaluation of the Destination Port Logistics based on Self-Organizing Map Computing. In 2022 IEEE 22nd International Conference on Software Quality, Reliability, and Security Companion (QRS-C) (pp. 746-751). IEEE.
[5] Ganesh, M. S. S., Sujith, M. B. R. P., Aravindh, K. V., & Durgadevi, P. (2023). Crime Prediction Using Machine Learning Algorithms. Advances in Science and Technology, 124, 457-461.
[6] Plevris, V., Solorzano, G., Bakas, N. P., & Ben Seghier, M. E. A. (2022). Investigation of performance metrics in regression analysis and machine learning-based prediction models. In 8th European Congress on Computational Methods in Applied Sciences and Engineering (ECCOMAS Congress 2022). European Community on Computational Methods in Applied Sciences.
[7] Yin, J. (2023). Crime Prediction Methods Based on Machine Learning: A Survey. Computers, Materials & Continua, 74(2).
[8] Alsayadi, H. A., Khodadadi, N., & Kumar, S. (2022). Improving the regression of communities and crime using ensemble of machine learning models. J. Artif. Intell. Metaheuristics, 1(1), 27-34.
[9] Campedelli, G. M. (2022). Machine Learning for Criminology and Crime Research: at the crossroads. Routledge.
[10] Herzog, C. (2022). On the risk of confusing interpretability with explicability. AI and Ethics, 2(1), 219-225.
[11] Ahamad, G. N., Shafiullah, Fatima, H., Imdadullah, Zakariya, S. M., Abbas, M., & Usman, M. (2023). Influence of optimal hyperparameters on the performance of machine learning algorithms for predicting heart disease. Processes, 11(3), 734.
[12] Omotehinwa, T. O., & Oyewola, D. O. (2023). Hyperparameter optimization of ensemble models for spam email detection. Applied Sciences, 13(3), 1971.
[13] Anyanwu, G. O., Nwakanma, C. I., Lee, J. M., & Kim, D. S. (2023). Novel hyper-tuned ensemble random forest algorithm for the detection of false basic safety messages in internet of vehicles. ICT Express, 9(1), 122-129.
[14] Wubineh, B. Z. (2024). Crime analysis and prediction using machine-learning approach in the case of Hossana Police Commission. Security Journal, 1-16.
[15] Wheeler, A. P., &Steenbeek, W. (2021). Mapping the risk terrain for crime using machine learning. Journal of Quantitative Criminology, 37, 445-480.
[16] Liao, M., Wen, H., Yang, L., Wang, G., Xiang, X., & Liang, X. (2024). Improving the model robustness of flood hazard mapping based on hyperparameter optimization of random forest. Expert Systems with Applications, 241, 122682.
[17] Watts, D., Moulden, H., Mamak, M., Upfold, C., Chaimowitz, G., & Kapczinski, F. (2021). Predicting offenses among individuals with psychiatric disorders-A machine learning approach. Journal of Psychiatric Research, 138, 146-154.
[18] Rodrigues, A., González, J. A., & Mateu, J. (2023). A conditional machine learning classification approach for spatio-temporal risk assessment of crime data. Stochastic Environmental Research and Risk Assessment, 37(7), 2815-2828.
[19] Abubakar, H., Boukari, S., Gital, A. Y., & Zambuk, F. U. (2024). A hyper-parameter tuned Random Forest algorithm-based on Artificial Bee Colony for improving accuracy, precision and interpretability of crime prediction. Dutse Journal of Pure and Applied Sciences, 10(4a), 371-381.
[20] Adorada, A., Permatasari, R., Wirawan, P. W., Wibowo, A., &Sujiwo, A. (2018, October). Support vector machine-recursive feature elimination (svm-rfe) for selection of Micro RNA expression features of breast cancer. In 2018 2nd international conference on informatics and computational sciences (ICICoS) (pp. 1-4). IEEE.
[21] Awad, M., &Fraihat, S. (2023). Recursive feature elimination with cross-validation with decision tree: Feature selection method for machine learning-based intrusion detection systems. Journal of Sensor and Actuator Networks, 12(5), 67.
[22] Van Patten, I. T., McKeldin-Coner, J., & Cox, D. (2009). A microspatial analysis of robbery: Prospective hot spotting in a small city. Crime Mapping: A journal of research and practice, 1(1), 7-32.
Cite This Article
Plain Text BibTeX RIS

  • APA Style

    Abubakar, H., Boukari, S., Gital, A. Y., Zambuk, F. U. (2025). Evaluation of a Hyperparameter Tuned Random Forest Algorithm Based on Artificial Bee Colony for Improving Accuracy and Precision of Crime Prediction Model. American Journal of Biological and Environmental Statistics, 11(3), 107-121. https://doi.org/10.11648/j.ajbes.20251103.16

    Copy | Download

    ACS Style

    Abubakar, H.; Boukari, S.; Gital, A. Y.; Zambuk, F. U. Evaluation of a Hyperparameter Tuned Random Forest Algorithm Based on Artificial Bee Colony for Improving Accuracy and Precision of Crime Prediction Model. Am. J. Biol. Environ. Stat. 2025, 11(3), 107-121. doi: 10.11648/j.ajbes.20251103.16

    Copy | Download

    AMA Style

    Abubakar H, Boukari S, Gital AY, Zambuk FU. Evaluation of a Hyperparameter Tuned Random Forest Algorithm Based on Artificial Bee Colony for Improving Accuracy and Precision of Crime Prediction Model. Am J Biol Environ Stat. 2025;11(3):107-121. doi: 10.11648/j.ajbes.20251103.16

    Copy | Download 

  • @article{10.11648/j.ajbes.20251103.16,
  author = {Hauwa Abubakar and Souley Boukari and Abdulsalam Ya’u Gital and Fatima Umar Zambuk},
  title = {Evaluation of a Hyperparameter Tuned Random Forest Algorithm Based on Artificial Bee Colony for Improving Accuracy and Precision of Crime Prediction Model
},
  journal = {American Journal of Biological and Environmental Statistics},
  volume = {11},
  number = {3},
  pages = {107-121},
  doi = {10.11648/j.ajbes.20251103.16},
  url = {https://doi.org/10.11648/j.ajbes.20251103.16},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ajbes.20251103.16},
  abstract = {Crime prediction is a crucial application of machine learning, enabling law enforcement to make proactive decisions. This research presents a novel crime prediction model that leverages a hybrid approach by tuning the hyperparameters of Random Forest (RF) classifier using the Artificial Bee Colony (ABC) optimization algorithm. The model was developed to improve the prediction accuracy and reliability of crime prediction systems by enhancing the performance of traditional machine learning classifiers. To validate the effectiveness of the proposed RF-ABC model, a comparative analysis against Decision Trees (DT), k-Nearest Neighbors (KNN), and the existing untuned Random Forest model was conducted. Experimental results demonstrate that the proposed RF-ABC model significantly outperforms the baseline models across multiple performance metrics. Specifically, the RF-ABC achieved an accuracy of 95%, precision of 90%, recall of 93%, and an F1-score of 90%. In comparison, the existing RF model yielded an accuracy of 81%, precision of 87%, recall of 84%, and an F1-score of 83%, while DT and KNN models recorded notably lower scores. DT obtained a PEI of 0.6900, PAI of 0.669 and RRI of 0.5200, while KNN has a PEI of 0.9647, PAI of 0.8670 and RRI of 0.5267, RF-ABC had the best result. PEI of 0.9800, PAI of 0.9000 and RRI of 0.7200. Crime prediction metrics show that These findings confirm that the integration of ABC with RF not only fine-tunes the hyperparameters efficiently but also enhances the model's predictive capabilities. The proposed hybrid approach shows promising potential for real-world crime analytic and decision support systems in law enforcement.
},
 year = {2025}
}

    Copy | Download 

  • TY  - JOUR
T1  - Evaluation of a Hyperparameter Tuned Random Forest Algorithm Based on Artificial Bee Colony for Improving Accuracy and Precision of Crime Prediction Model

AU  - Hauwa Abubakar
AU  - Souley Boukari
AU  - Abdulsalam Ya’u Gital
AU  - Fatima Umar Zambuk
Y1  - 2025/09/11
PY  - 2025
N1  - https://doi.org/10.11648/j.ajbes.20251103.16
DO  - 10.11648/j.ajbes.20251103.16
T2  - American Journal of Biological and Environmental Statistics
JF  - American Journal of Biological and Environmental Statistics
JO  - American Journal of Biological and Environmental Statistics
SP  - 107
EP  - 121
PB  - Science Publishing Group
SN  - 2471-979X
UR  - https://doi.org/10.11648/j.ajbes.20251103.16
AB  - Crime prediction is a crucial application of machine learning, enabling law enforcement to make proactive decisions. This research presents a novel crime prediction model that leverages a hybrid approach by tuning the hyperparameters of Random Forest (RF) classifier using the Artificial Bee Colony (ABC) optimization algorithm. The model was developed to improve the prediction accuracy and reliability of crime prediction systems by enhancing the performance of traditional machine learning classifiers. To validate the effectiveness of the proposed RF-ABC model, a comparative analysis against Decision Trees (DT), k-Nearest Neighbors (KNN), and the existing untuned Random Forest model was conducted. Experimental results demonstrate that the proposed RF-ABC model significantly outperforms the baseline models across multiple performance metrics. Specifically, the RF-ABC achieved an accuracy of 95%, precision of 90%, recall of 93%, and an F1-score of 90%. In comparison, the existing RF model yielded an accuracy of 81%, precision of 87%, recall of 84%, and an F1-score of 83%, while DT and KNN models recorded notably lower scores. DT obtained a PEI of 0.6900, PAI of 0.669 and RRI of 0.5200, while KNN has a PEI of 0.9647, PAI of 0.8670 and RRI of 0.5267, RF-ABC had the best result. PEI of 0.9800, PAI of 0.9000 and RRI of 0.7200. Crime prediction metrics show that These findings confirm that the integration of ABC with RF not only fine-tunes the hyperparameters efficiently but also enhances the model's predictive capabilities. The proposed hybrid approach shows promising potential for real-world crime analytic and decision support systems in law enforcement.

VL  - 11
IS  - 3
ER  -

    Copy | Download

Author Information

  • Hauwa Abubakar

    Computer Science Department, Abubakar Tafawa Balewa University Bauchi, Bauchi, Nigeria

    Biography: Hauwa Abubakar is a PhD Student at Abubakar Tafawa Balewa University Bauchi (ATBU), Bauchi, Nigeria. She Obtained a Bachelor’s degree of Technology and a Master’s of Science in computer science from ATBU. She has attended a number of local and international Conferences in Nigeria. Her current research areas includes, Network security, machine learning, and cyber security

    Contact Email

    http://orcid.org/0000-0002-6341-4169

  • Souley Boukari

    Computer Science Department, Abubakar Tafawa Balewa University Bauchi, Bauchi, Nigeria

    Contact Email

    http://orcid.org/0000-0002-8653-4083

  • Abdulsalam Ya’u Gital

    Computer Science Department, Abubakar Tafawa Balewa University Bauchi, Bauchi, Nigeria

    Biography: Abdulsalam Ya’u Gital is a Professor at Abubakar Tafawa Balewa University Bauchi (ATBU), Bauchi, Nigeria. He is currently involved in teaching, research and supervision of both undergraduate and Postgraduate students in computer science. he has a B-Tech, from ATBU, Bauchi Nigeria, M. Sc. and PhD in computer science from the University of Technology Malaysia (UTM). he serves as the Director of ICT and Deputy Dean of the Faculty of Science at ATBU. A professional member of several prestigious organizations. Professor Gital’s affiliations include the Association for Computing Machinery (ACM), IEEE computer society, Nigerian Computer Society (NCS), Computer Professional Registration Council of Nigeria (CPN) and Teachers without Borders African Regional chapter. His research interest encompasses, Deep Learning, Cloud Computing, Intelligent Systems and the Internet of Things (Io T). His work has been widely recognized in international referred journals and conferences highlighting his significant contributions to the field.

    Contact Email

  • Fatima Umar Zambuk

    Computer Science Department, Abubakar Tafawa Balewa University Bauchi, Bauchi, Nigeria

    Biography: Fatima Umar Zambuk is currently serving as the Head of the Department Computer Science, Faculty of Computing Abubakar Tafawa Balewa University Bauchi (ATBU), Bauchi, Nigeria. She Obtained a Bachelor’s Degree of Technology, a Master’s of Science and PhD in Computer Science from ATBU. She is currently involved in teaching and supervision of both Undergraduate and Postgraduate students of Computer Science from ATBU. Her current research areas includes, Cloud Computing, machine learning, and Optimization.

    Contact Email

    http://orcid.org/0000-0003-4276-4490

Download PDF
Submit an Article
  • Abstract
  • Keywords

  • Document Sections

    1. 1. Introduction
    2. 2. Related Works
    3. 3. Materials and Methods
    4. 4. Results
    5. 5. Discussions
    6. 6. Conclusion
    Show Full Outline
  • Abbreviations
  • Author Contributions
  • Funding
  • Data Availability Statement
  • Conflicts of Interest
  • References
  • Cite This Article
  • Author Information

About Us

Science Publishing Group (SciencePG) is an Open Access publisher, with more than 300 online, peer-reviewed journals covering a wide range of academic disciplines.

Learn More About SciencePG

Products

Information

Important Link

Copyright © 2012 -- 2025 Science Publishing Group – All rights reserved.