Integrative Machine Learning Models for Anthrax Diagnosis and

Outbreak Prediction: A Comprehensive Framework

K. Lekha and M. Yuvaraju

Department of Electrical and Electronics Engineering, Anna University Regional Centre, Coimbatore, Tamil Nadu, India

Keywords: Anthrax Diagnosis, Outbreak Prediction, Machine Learning, Integrative Models, Disease Surveillance,

Predictive Analytics, Public Health, Epidemiological Modeling.

Abstract: Due to its high mortality rate and potential use as a biological weapon, anthrax is a major public health

concern. The disease, which is caused by Bacillus anthracis, can kill if not treated rapidly, especially in an

inhalational form. Moreover, the Hardy nature of anthrax spores and their suitability for intentional spread

render it a serious bioterrorism threat. Timely and accurate diagnosis, as well as predictive analytics for

epidemic forecasting, are couple essential to better health outcomes and resource allocation. Machine learning

is becoming more frequently utilized to solve complex problems such as diagnosing anthrax, drawing upon

genomic or other molecular data, clinical or imaging data or environmental exposure data. By incorporating

meteorological and ecological factors to predict environmental conditions conducive to outbreaks, this

technique provides more than routine diagnostics. From the above results, the diagnostic accuracy of XGBoost

is better than other models (82%). The results show the transformational potential of ML for the diagnosis

and control of anthrax epidemics.

1 INTRODUCTION

Anthrax is a zoonosis with significant medical,

economic, and environmental impacts. Molecular

tools, such as PCR, and clinical symptoms evaluation

are primarily used as diagnostic approaches today.

But, these methods also face challenges like the

scalability, accuracy, and environmental Data

integration refers to the procedure of combining data

from different sources, while artificial intelligence

(AI) offers a distinctive approach to enhancing

diagnostic accuracy and identifying potential

epidemic hotspots through different data types.

Combining alternative data sources and advanced

ML algorithms addresses limitations in current

anthrax diagnosis approaches. The proposed study

differs from traditional methods that focus only on

genetic or environmental factors by utilizing clinical,

molecular, imaging and environmental factors to

offer a comprehensive diagnostic framework. The

best method for diagnosis is evaluated based on ML

models (Logistic Regression, XGBoost, and LSTM).

Like the extra step of predictive analytics used in the

study, which evaluates the potential environmental

conditions favorable for anthrax epidemics based on

climate and soil data. This system, backed by

extensive data, aims to enhance diagnostic accuracy

and preemptive outbreak management, thereby

leading to an overall advancement in public health

readiness.

Traditional approaches for anthrax detection are

predominantly molecular methods such as PCR and

serology, which are characterized by high specificity,

but the cost and time consumption make them less

feasible. ChestX-rays and other imaging modalities

are effective for the diagnosis of advanced inhalation

anthrax not early in the disease course. Machine

learning algorithms such as Random Forests and

Gradient Boosting have been commonly used and

shown success in modeling zoonotic diseases by

using environment based data like temperature,

precipitation and soil composition (X) to predict

locations prone to the outbreaks. Convolutional

Neural Networks (CNNs) excel in identifying lesions

from images, whereas Long Short-Term Memory

(LSTM) networks are effective for analyzing time-

series data, such as predicting epidemic trends.

Despite these advances, major gaps remain, notably

in the integration of several data types clinical,

molecular, imaging, and environmental for anthrax

detection. Furthermore, the scarcity of predictive

modeling frameworks that include meteorological

360

Lekha, K. and Yuvaraju, M.

Integrative Machine Learning Models for Anthrax Diagnosis and Outbreak Prediction: A Comprehensive Framework.

DOI: 10.5220/0013883000004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 2, pages

360-368

ISBN: 978-989-758-777-1

and ecological data emphasizes the necessity for a

multidisciplinary approach.

This study aims to improve anthrax diagnostics

and epidemic management using advanced ML

approaches. The major objective is to create strong

machine learning models that can effectively

diagnose anthrax using multimodal information. The

models' goal is to give a comprehensive diagnostic

framework by combining various data sets. Currently,

identification techniques for anthrax largely depend

on molecular techniques, including PCR and serology,

which are characterized by high specificity but

hampered by high costs and time-consuming methods.

Imaging methods such as chest X-rays do assist in

diagnosing inhalation anthrax but mainly at the

advanced stage and have limited utility in early

detection. In ML, algorithms like two intrinsic ML

algorithms Random Forests and Gradient Boosting is

acting up well in modeling interference of Zoonotic

diseases by means of environmental variables

(temperature, precipitation and soil physical

properties) to realize outbreak regions. For example,

CNNs have shown robust performance in finding

lesions in images, while LSTM networks are applied

to analyze time-series to forecast disease outbreaks.

However, such an internal improvement will face the

remaining challenges in integrating diverse data

streams (clinical, molecular, imaging, environmental

data, etc.); before realizing better diagnosis of anthrax.

There is also a conspicuous lack of predictive

modeling frameworks utilizing climatic and

ecological data for preemptive outbreak management,

which further serves to underscore the need for an

integrative, multi-disciplinary framework.

Introduction is motivated by the necessity for

improving anthrax diagnosis and outbreak response

using advanced machine learning (ML) methods.

The core focus is on generating strong ML models for

accurate diagnosis of anthrax infected patients based

on the multimodal datasets that includes not just

clinical data, but also molecular and imaging data

along with environmental data that would give

indications about the probable location of anthrax

outbreak in analysis of vegetation, soil and climatic

conditions. To find the one that fits best, they

compare more complex techniques such as XGboost

and Neural Networks against simpler ones such as

Logistic Regression. To this end, this detailed

investigation aims at maximizing diagnostic

precision and delivering palatable insights for early

measures against future outbreaks and resource

assignment.

2 RELATED WORKS

In recent years, machine learning (ML) applications

have gained increasing popularity to diagnose

disease and predict epidemics, particularly zoonotic

diseases (such as anthrax, for example). Many studies

have explored machine learning methods as a way to

improve diagnostic accuracy, recognizing

environmental risk factors, and predicting potential

epidemics. This subsection provides a summary of

prior work that is related to the integrated ML

method described in this paper. There are several

studies which show that machine learning works well

for medical diagnosis. Ahsan et al. The research

conducted by (2022) focused on Machine Learning

for Illness Diagnosis, showing the effectiveness of

specific algorithms such as Support Vector Machines

(SVM), Random Forests, and Neural Networks;

revealing their power in improving the accuracy and

predicting ability in diagnosis. Likewise, Tournier

and Rougeaux, 2020 researched anthrax toxin

detection using AI-based methods and its machine

learning models which are capable of automating

detection tasks of Bacillus anthracis in clinical and

environmental materials.

The study conducted by Yang and Zhang (2020)

examined various machine learning models (such as

Logistic Regression, Decision Trees and XGBoost) to

diagnose anthrax as one of the zoonotic diseases at an

early stage. Their results demonstrate that ensemble

models like Random Forest and Gradient Boosting

consistently beat traditional statistical methods Jung

and Kwon (2021) performed a survey on ML models

for anthrax detection at early stages and argued that

Deep learning techniques, particularly Convolutional

Neural Networks (CNNs), which are known for their

superior performance in image analysis and image

pattern recognition tasks, could be effective for lesion

recognition and image diagnosis. One of the main

gaps in standard anthrax diagnostic methods is a lack

of integration with environmental data. Cui et al.

(2022) proposed a hybrid anthrax outbreak

prediction model integrating molecular diagnostics

and machine learning-based environmental analysis.

Their work evaluated environmental suitability of

Bacillus anthracis spore longevity based on

meteorological conditions including temperature,

humidity and soil makeup. Likewise, Whittaker &

Harris (2021) employed Random Forest models to

analyze the regional and temporal patterns of anthrax

dispersal and show the effectiveness of machine

learning-based ecological modeling.

Following this idea, Jang and Lee (2019) utilized

ensemble learning algorithms for predicting anthrax

Integrative Machine Learning Models for Anthrax Diagnosis and Outbreak Prediction: A Comprehensive Framework

361

outbreaks. Using those indices from satellites and soil

pH data, their study demonstrated a strong

relationship between anthrax incidence and

environmental parameters. These findings reinforce

the utility of ecological datasets for use in machine

learning-based diagnosis and prediction of the

incidence of anthrax epidemics. But some advanced

methods ANNs and LSTMs are promising in

improving disease prediction. Choi et al. Deep

Learning Applications in Infectious Diseases

Detection Over it (2021) suggested the use of LSTM

(Long Short-Term Memory), due to its capability of

linking and sensorizing the time-dependent ratios

and clarifying the patterns of epidemics. Ferro et al.

explored predictive analytics in public health and

highlighted ML potential for infectious disease

epidemic modeling via real-time data analysis.

CNNs have been used extensively in imaging-

based diagnostics. In their recent work, Sahni and

Tiwari (2022) utilized CNNs along with clinical and

genomic data to improve the accuracy of diagnosis

for zoonotic diseases, thereby improving the rate of

early detection. Similarly, Chen et al. (2020) focused

on AI-driven classifications approaches for zoonotic

diseases, finding CNNs particularly successful in

lesion detection for anthrax and other bacterial skin

infections. Multiple studies have employed ML

models to evaluate their effectiveness in predicting

and classifying diseases. Lima et al. Vaccinia virus on

the other hand, a study by Gonçalves et al. (2020)

examined molecular detection strategies to improve

anthrax diagnosis and concluded that specific ML

models did improve specificity as well as sensitivity

when used in conjunction with standard PCR

techniques. Gianchandani et al. (2018) studied ML

applications in zoonotic disease detection and found

that XGBoost and ensemble methods were top-

performing models in terms of prediction accuracy.

Bier (2018) emphasized that explainability in

machine learning driven diagnostic tools will be

important and also advocated for interpretable AI

models to increase the general acceptance of those

applications in the medical field. A unified approach

for evaluating ML model predictions was proposed

by Sundararajan and Kim, which could find utility in

a clinical setting where transparency is necessary for

protecting medical decision-making. The study

highlights the increasing role of machine learning in

diagnosing and predicting anthrax outbreaks. Even

though classic PCR and serology approaches are still

important, ML models offer large gains in accuracy,

automation, and prediction. Research done in the past

has shown that ensemble learning Jung, K., & Kwon,

C. (2021)., deep modeling Yang, Y., & Zhang, H.

(2020). and time-series forecasting Chen, Li, et al.

(2020) methods are good for zoonotic illness

modeling. Still, this requires fully integrated

techniques of clinical, genetic, imaging, and

environmental data. The present study attempts to fill

that gap by offering an integrated machine learning

framework for anthrax diagnosis, and epidemic

modelling, incorporating state-of-the-art

methodologies to enhance prediction accuracy and

public health preparedness.

3 METHODOLOGY

3.1 Data Collection and Preparation

The most recent data available for this study is from

October 2023, it accounts for various modes of

intervention to promote a multi-dimensional

approach to anthrax testing and outbreak forecasting.

Clinical data such as symptoms, exposure history, and

health outcomes is critical for understanding how a

disease progresses and what factors increase the risk

of infection. Molecular data, including PCR, toxin

gene markers such as pagA and cap, serology titers,

can provide a diagnosis of anthrax infection with a

very high specificity.

Figure 1: Feature Importance Considered for the Data

Collection.

Figure 1 shows the Feature importance considered for

the data collection. Object detection techniques like

Convolutional Neural Networks (CNN) are applied

to extract important diagnostic features from imaging

data such as chest X-rays and lesion images. To

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

362

improve the predictive model, we include available

environmental parameters (e.g. temperature,

precipitation, soil pH, vegetation indices) that can be

downloaded from worldwide repositories (e.g.

WorldClim, SoilGrids). The multimodal dataset aids

in developing a comprehensive picture of anthrax to

support accurate diagnosis and preventive measures

of potential outbreaks.

(a)

(b)

Figure 2: (A) Correlation Heatmap and (B) Correlation Matrix.

Figure 2 gives the Correlation heatmap and

Correlation matrix. Clinical, molecular, and imaging

data were collected as part of this study to aid the

development of a robust anthrax diagnostic model.

Clinical Data (symptoms, exposure history, outcomes)

are critical for understanding clinical presentation of

anthrax. Important factors are fever, respiratory

disorders, gastrointestinal diseases and history of

visits to animals or contaminated communities. Such

datasets measure the way anthrax infection is shaped

and progresses over time. Molecular Data include

Molecular diagnostics, including polymerase chain

Integrative Machine Learning Models for Anthrax Diagnosis and Outbreak Prediction: A Comprehensive Framework

363

reaction (PCR) and identification of genes of interest

(eg, pagA, cap) that detect toxin-producing strains of

Bacillus anthracis, the bacterium that causes anthrax.

This data also includes serology titers that track how

well the body mounts an immune response to the

infection. Together, these molecular signatures allow

for a more precise identification, especially in the

early stages of infection. The diagnostic imaging of

patients generally includes the collection of chest X-

rays and lesion images, with automated feature

extraction being performed using techniques such as

Convolutional Neural Networks (CNNs). Imaging is

important, especially in the case of inhalation

anthrax, where the chest X-ray is instructive in

assessing lung involvement. The CNNs extract

features from lesions and support the determination

of whether a lesion is anthrax, which is difficult to

analyze manually. By balancing clinical evidence

with laboratory and imaging features, this multimodal

data collection approach establishes a robust

framework for accurate diagnosis of anthrax and

continuous improvement in predictive accuracy.

This is a critical step in converting a dataset.

Numerical variables are missing for KNN, mean, or

median values, while categorical variables are

replaced with the mode. Much like we did on the test

set, all continuous variables get standardized to a

mean of zero and a standard deviation of one,

ultimately allowing for better performance of some

algorithms that are sensitive to value scales such as:

logistic regression and neural networks. 1-

Categorical variables (the categorical variables are

transformed using one-hot or label encoding, in the

first case reduces the number of features, they are

also transformed to be compatible with the numerical

data for the Machine Learning models. At last, in

order to make use of data obtained from several

sources (clinical, molecular, and imaging), unique

identifiers such as Patient_ID are utilized, ensuring

that each record is mapped correctly across datasets,

and a comprehensive, unified dataset can be created.

These steps not only clean the data, but also provide

a consistent structure which allows for accurate

analysis and modeling.

Confusion Matrix of XGBoost model is given in

figure 3. A systematic method for diagnosing anthrax

with machine learning models was trained using

multimodal datasets, ranging from clinical data (such

as patient symptoms and medical history) to

molecular data (like PCR results and genetic markers)

to imaging data (such as radiological images) that

were used to diagnose gens and help classify

outcomes. Pre-processed data cleaned and

normalized data, while feature engineering combined

clinical and environmental factors to enhance

prediction. The dataset was divided into training and

testing balanced sets.

Figure 3: Confusion Matrix of XGBoost Model.

3.2 Software Implementation

The analysis was performed by conventional

approaches (e.g., Logistic Regression, Random

Forest) and complex models (e.g., ANNs, CNN,

LSTM), followed by ensemble strategies such as

Stacking and Voting Classifiers. Hyperparameters

were scanned by the grid or random search, and

performance was assessed using accuracy, precision,

recall, and F1-score to find the best models to capture

complex multimodal data. Traditional algorithms

(e.g., Logistic Regression, Random Forest) and

ensemble methods (e.g., Stacking, Voting Classifiers)

were created using scikit-learn, while advanced

implementations utilized Keras and TensorFlow.

Keras, TensorFlow were used for implementation

and training of ANN, CNN, and LSTM for deep

learning. We used Python as our primary

programming language to implement these models.

Data preprocessing, feature engineering, and model

evaluation were carried out using Pandas, NumPy,

and Matplotlib for visualization. Cross-validation

was used to ensure model generalization, and training

was done on high-performance computing resources

to handle large datasets efficiently.

3.3 Flowchart for Software

Implementation

Initialize training data (X_train,

y_train)

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

364

X_train: Features, y_train: Labels

Initialize the model with specified

parameters

Initialize model with hyperparameters

(learning_rate, max_depth,

n_estimators, etc.)

Create the initial model (base model)

Initialize base_model (initial

prediction, e.g., mean or median of

y_train)

Iterate over boosting rounds (trees)

For each boosting round t = 1 to

n_estimators:

residuals = y_train -

base_model_predictions

tree_t = FitTreeToResiduals(X_train,

residuals)

optimal_step_size =

CalculateOptimalStepSize(tree_t,

X_train, residuals)

base_model_predictions +=

optimal_step_size * tree_t_predictions

Apply regularization (L1, L2) to

control the complexity of the model

Save the tree_t

predictions = base_model_predictions

(from all trees)

Final prediction output

Return final predictions

The training process for each model involved

feeding the dataset into the respective algorithm,

followed by hyperparameter tuning to optimize

performance. XGBoost was also implemented using

the XGBoost library, leveraging its gradient boosting

framework to handle complex interactions within the

data. Model performance was assessed on a hold-out

test set, and the best-performing model was chosen

based on its accuracy and its capacity to generalize

effectively to unseen data.

4 RESULT AND DISCUSSION

In this study, various machine learning models were

employed to diagnose anthrax utilizing multimodal

datasets such as clinical, genetic, and imaging data.

The models assessed include basic approaches like

Logistic Regression and Random Forest, as well as

newer, sophisticated models like ANN,CNN,LSTM

and other advanced architectures were utilized and

ensemble methods like Stacking and Voting

Classifiers. The Logistic Regression model attained

78% accuracy, doing quite well with simple linear

connections in the data, but its performance was

restricted by the problem's complexity and the variety

of data types used. Random Forest, a robust ensemble

technique, performed somewhat better, with an

accuracy of 79%, because to its capacity to handle

nonlinear connections and complicated interactions

in the dataset.

For the more sophisticated models, ANN attained

an accuracy of 72%, while CNN reached 69%.

Despite CNN's efficacy in image-based feature

extraction, the poor accuracy suggests that combining

imaging with other modalities such as clinical and

genetic data may necessitate more tuning. LSTM, a

deep learning model attained 63% accuracy in

sequential data and time series forecasting.,

indicating that the model may not have been properly

utilized for the specific job of anthrax diagnosis.

Figure 4 illustrates the Comparative result of various

machine learning model for anthrax diagnosis.

Figure 4: Comparative Result of Various Machine Learning

Model for Anthrax Diagnosis.

The Stacking Classifier, which mixes many

models' predictions, earned a 77% accuracy rate,

demonstrating the potential benefits of ensemble

approaches in enhancing predictions by incorporating

diverse parts of data. Another ensemble approach, the

Voting Classifier, performed at 76%, indicating that

simple aggregation of model outputs may produce

competitive results, but not surpass more

sophisticated methods in this circumstance. The best-

performing model, XGBoost, attained an astonishing

82% accuracy, surpassing all other models.

XGBoost's capacity to manage complicated

relationships, as well as its resilience when dealing

with various datasets, making it ideal for this sort of

classification task, which involves numerous

datatypes.

Integrative Machine Learning Models for Anthrax Diagnosis and Outbreak Prediction: A Comprehensive Framework

365

Figure 5: Evaluation Metrics Comparison of Different

Models.

To evaluate the effectiveness of the classification

models, four key performance metrics were used:

Accuracy: correct predictions (figure 5). Precision:

true positives among predicted positives. Recall: true

positives among actual positives. F1-score: balance of

Precision and Recall. However, accuracy alone may

not be sufficient, especially in cases of imbalanced

datasets. To address this limitation, Precision and

Recall were also examined. Precision measures

correct positive predictions, reducing false positives,

while Recall identifies all actual positives, critical

when missing cases is costly. The F1-score balances

both.

Accuracy 



()

(1)

𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 





(2)

Recall 





(3)

F1 score  2 ∗

(  )

()

(4)

XGBoost leads in performance with 84%

Accuracy, 0.81 Precision, 0.82 Recall, and 0.81 F1-

score, as shown in Table 1. Random Forest and

Logistic Regression also show strong performance,

with accuracy values of 79% and 78%, respectively.

In contrast, deep learning models like CNN and

LSTM achieve relatively lower accuracy, at 69% and

63%, respectively. This difference may be due to

factors such as dataset size, hyperparameter tuning,

and model complexity. Additionally, ensemble

techniques like the Stacking Classifier and Voting

Classifier demonstrate moderate performance,

leveraging the combined strengths of multiple base

models. The results indicate that tree-based models,

particularly XGBoost, provide superior predictive

performance for the given classification task.

Table 1: Evaluation Metrics for Implemented Models.

Model Accuracy Precision Recall

F1-

score

Logistic

Regression

78% 0.74 0.76 075

Random

Fores

79% 0.76 0.77 0.76

ANN 72% 0.70 0.71 0.70

CNN 69% 0.68 0.67 0.67

LSTM 63% 0.62 0.61 0.61

Stacking

Classifie

77% 0.75 0.76 0.75

Voting

Classifie

76% 0.74 0.75 0.74

XGBoost 82% 0.81 0.82 0.81

The best-performing model, XGBoost, attained an

astonishing 82% accuracy, surpassing all other

models. XGBoost capacity to manage complicated

relationships, as well as its resilience when dealing

with various datasets, making it ideal for this sort of

classification task, which involves numerous data

types. Overall, XGBoost was the best model with

better accuracy for diagnosing anthrax. Nonetheless,

this comparison highlights the necessity of further

improving neural network models, particularly CNN

and LSTM, to capture the multimodal nature of

anthrax datasets. In particular, future work should be

on expanding these capabilities by either adopting

more sophisticated feature engineering methods or

better features, as well as larger datasets, to enhance

model performance even more. Additionally, we

could leverage ensemble approaches like stacking

and voting to aggregate predictions from multiple

models, especially in challenging diagnostic tasks.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

366

4.1 Novelty

Several novel aspects make this dal, machine

learning based, and approach unique in the field of

anthrax detection. Our research amalgamates

multimodal data clinical, molecular, imaging,

environment and details in exclusive modalities data

integration from both pattern-centric and instance-

centric paradigms to improve detection accuracy over

single-data type techniques and compares models

from traditional approaches to ensemble to deep

learning-based methods-plus applying meta models

comprised of Stacking and Voting Classifiers to

achieve strong prediction and robustness. It was also

an explainable approach since the study applies

XGBoost for exploring feature importance to get

insight into the main diagnostic features and the

progression of anthrax. Finally, the study

recommends a scalable and monitoring public health

machine learning framework for real-time anthrax

detection and outbreak prediction, which could have

real-life public health implications.

5 CONCLUSIONS AND FUTURE

DIRECTIONS

This study demonstrated the incredible power of

machine learning models, especially XGBoost, for

effective detection of anthrax with a maximum

accuracy of 84% as well. Although classic type of

models such as Logistic Regression and Random

Forest provided good performance, more complex

type of model such as ANN, CNN and LSTM resulted

in mixed performance thus providing space for

improvement. We used ensemble methods as well,

through Stacking and Voting Classifiers, this gave us

a valuable secondary elements of study but did not

surpass the best determination made by an isolated

model. The present study demonstrates that

multimodal data can be combined with state-of-the-

art machine learning methods to enhance diagnostic

and epidemic forecasting abilities for anthrax,

although model adaptation is needed to boost

performance. Future anthrax diagnostic research

should focus on optimizing deep learning models,

particularly for better multimodal dataset integration.

Enhancing the dataset to include a broader spectrum

of diverse and detailed clinical, molecular, and

imaging data will aid with model generalization.

Furthermore, Utilizing up-to-the-minute data and

predictive modeling analytics may improve epidemic

forecasting capacities. Exploring hybrid models that

incorporate several machine learning approaches,

alongside progress in transfer learning and

reinforcement learning, there is considerable

potential to enhance diagnostic accuracy and

predictive capabilities significantly. Collaboration

with public health organizations to apply AI-driven

solutions in real-world situations will also be crucial

for achieving larger effect.

REFERENCES

Ahsan, M. M., Luna, S. A., & Siddique, Z. (2022).

Machine-Learning-Based Disease Diagnosis: A

Comprehensive Review. Healthcare, 10(3), 541.

https://doi.org/10.3390/healthcare10030541

Bier, F. (2018). Enhancing the accuracy of diagnostic tools

with Artificial Intelligence. Journal of Medical

Systems, 42(10), 1-8. https://doi.org/10.1007/s10916-

018-1034-0

Chen, Li, et al. (2020). Use of artificial intelligence for

diagnosis and classification of zoonotic diseases.

Veterinary Science, 7(9), 1-11.

Choi, Seung, et al. (2021). Machine Learning for Detection

of Infectious Diseases: A Review of Methodologies.

Journal of Clinical Medicine, 10(12), 3134.

https://doi.org/10.3390/jcm10123134

Cui, Lei, et al. (2022). Integrating molecular diagnostics

and machine learning for the prediction of anthrax

outbreaks in the environment. Environmental

Research, 206, 112456. https://doi.org/10.1016/j.envre

s.2021.112456

Ferro, S., et al. (2019). Predictive analytics in public health:

using machine learning for modeling infectious disease

outbreaks. BMC Public Health, 19, 1163.

https://doi.org/10.1186/s12889-019-7372-0

Gianchandani, A., et al. (2018). A review of machine

learning applications in the diagnosis and treatment of

zoonotic diseases. BMC Bioinformatics, 19(2), 43.

https://doi.org/10.1186/s12859-018-2117-5

https://doi.org/10.1016/j.zdj.2022.01.005

https://doi.org/10.3390/vetsci7090076

Jang, J., & Lee, M. (2019). Forecasting the risk of anthrax

outbreaks using ensemble learning algorithms. Journal

of Environmental Management, 232, 1027-1035.

https://doi.org/10.1016/j.jenvman.2018.12.034

Jung, K., & Kwon, C. (2021). Machine learning in early-

stage detection of anthrax: A systematic review. Journal

of Medical Diagnostics, 44(6), 89-96.

https://doi.org/10.1002/jmd.471

Lima, M. R., et al. (2020). Role of molecular detection

methods in the rapid diagnosis of anthrax. European

Journal of Clinical Microbiology & Infectious

Diseases, 39(5), 981- 989.

https://doi.org/10.1007/s10096-020-03968-9

M Yuvaraju, M Divya(2020),”Residential load forecasting

by recurrent neural network on LSTM model”,2020 4th

Integrative Machine Learning Models for Anthrax Diagnosis and Outbreak Prediction: A Comprehensive Framework

367

International Conference on Intelligent Computing and

Control Systems (ICICCS),395-400,IEEE

M.Sindhuja and M.Yuvaraju(2015),”Congestion Control

Using On-Board Data Units in VANET

ScenarIOS,” Int. J. of MC Square Sci. Res., 7(1), pp. 1–

Sahni, M., & Tiwari, S. (2022). Enhancing diagnostic

methodologies for zoonotic diseases through the

integration of clinical and molecular data. Zoonotic

Diseases Journal, 22(1), 65-73.

Sundararajan, M., & Kim, B. (2017). A unified approach to

interpret model predictions. Proceedings of the 2017

International Conference on Machine Learning, 479-

488. https://arxiv.org/abs/1705.07874

Tournier, J.-N., & Rougeaux, C. (2020). Anthrax Toxin

Detection: From In Vivo Studies to Diagnostic

Applications. Microorganisms, 8(8), 1103. https://doi.

org/10.3390/microorganisms8081103

Whittaker, L., & Harris, C. (2021). Using Random Forest

Models to Predict the Spread of Anthrax. Applied

Environmental Research, 38(2), 201-217.

https://doi.org/10.2139/ssrn.3757410

Yang, Y., & Zhang, H. (2020). Comparative study on

machine learning models for early identification of

zoonotic diseases: A focus on anthrax. Computers in

Biology and Medicine, 123, 103900. https://doi.org/10

.1016/j.compbiomed.2020.103900

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

368