Prediction of Sepsis Using Light Gradient-Boosting Machine

Classifier in Comparison with Adaboost Classifier Based on Accuracy

Chindukuru Naga Sai Sreedhar and Loganayagi S.

Department of Computer Science and Engineering, Saveetha School of Engineering,

Saveetha Institute of Medical and Technical Sciences, Chennai, Tamil Nadu, 602105, India

KeywoRds: Sepsis, Machine Learning, Adaboost Classifier, Innovative Novel LightGBM Technique, Clinical Data,

Health.

Abstract: This study introduces a method to forecast sepsis employing the innovative LightGBM classifier model,

juxtaposing its improved accuracy against the Adaboost Classifier model. The dataset was sourced from

PhysioNet/Computing in Cardiology Challenge 2019's training set. The G power software informed the

sample size decision, suggesting 10 participants for each group, adopting a pretest power of 80%. A 95%

confidence interval was applied, and a significance level was established at 0.05%. Remarkably, the

LightGBM Technique achieved 96.41% accuracy, surpassing the AdaBoost Classifier's 77.58%. A significant

difference was observed between the two, evidenced by a P value of 0.019. In conclusion, the Light Gradient-

Boosting Machine classifier offers superior accuracy in predicting sepsis events.

1 INTRODUCTION

Sepsis is a severe condition that can become fatal

when the body overreacts to an infection. It can affect

people of all ages and may be triggered by various

infections, including bacterial, viral, and fungal

(Pravda, 2021). An exaggerated immune response to

these infections can damage healthy tissue. Using

machine learning for early sepsis detection is crucial

as it can enhance patient outcomes and potentially

save lives (Liu et al., 2019; G. Ramkumar et al.,

2021). Diagnosing sepsis can be challenging due to

its varied symptoms, which can resemble other health

issues. Progress in sepsis prediction through clinical

data analysis can help identify patients at high risk of

developing sepsis, facilitating quicker treatment

initiation (Reyna et al., 2019). Improved sepsis

prediction can not only better patient outcomes but

also facilitate bespoke treatment plans based on

individual risk (Wong et al., 2015). The research's

applications involve machine learning algorithms

capable of identifying patients at high risk for sepsis

during clinical trials, encouraging early intervention,

enhancing outcomes, and bolstering overall health

(Deepak et al., 2020; Sivakumar et al., 2022). These

algorithms can also craft predictive models gauging

sepsis probability using factors like demographic

data, health histories, and laboratory results.

Healthcare providers can then assess the situation

using the proposed model. Numerous research

articles have been published on sepsis prediction due

to its grave implications. Over 60 research articles on

sepsis prediction are available on ResearchGate, and

75 are found on Google Scholar. In a study by Taylor

et al. (2016), a random forest model was employed,

and its efficacy was compared with a classification

and regression tree (CART) model and a logistic

regression model. The random forest model

registered an AUC of 0.86 with a 95% confidence

interval, whereas the CART and logistic regression

models recorded AUCs of 0.69 and 0.76,

respectively. Bloch et al. (2019) reported that the

pooled area under the receiving operating curve

(SAUROC) for predicting sepsis 3 to 4 hours prior to

onset was 0.89, while the pooled SAUROCs for

SIRS, MEWS, and SOFA stood at 0.70, 0.50, and

0.78, respectively. In their research, Shrestha et al.

(2021) recommended a solution employing Gradient

Boosting that achieved a classification accuracy of

97.67%, compared to the typical 91.12% accuracy. X.

Peng and his team in their 2018 paper suggested a

methodology using a mixture-of-experts framework

to individualise sepsis treatment. This model

selectively oscillated between kernel (neighbour-

based) and DRL (Deep reinforcement learning)

methods. The paper by Shrestha et al. (2021) remains

Sreedhar, C. and S., L.

Prediction of Sepsis Using Light Gradient-Boosting Machine Classiﬁer in Comparison with Adaboost Classiﬁer Based on Accuracy.

DOI: 10.5220/0012570200003739

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

In Proceedings of the 1st International Conference on Artiﬁcial Intelligence for Internet of Things: Accelerating Innovation in Industry and Consumer Electronics (AI4IoT 2023), pages 595-600

ISBN: 978-989-758-661-3

595

the pinnacle in this area, with its model performance

being considerably superior to others.

Machine learning algorithms' efficacy heavily

hinges on the quality of the clinical data they're

trained upon. Should this data be skewed or

incomplete, the algorithm's performance could suffer,

highlighting a research gap. The research thus

discussed methods for deriving balanced data from an

imbalanced dataset. This research's objective is to

predict sepsis using the innovative novel LightGBM

classifier model and contrast its accuracy against the

Adaboost classifier.

2 MATERIALS AND METHODS

The preliminary research was undertaken at the

Machine Learning laboratory of Saveetha School of

Engineering, affiliated with the Saveetha Institute of

Medical and Technical Sciences, situated in Chennai.

The study used two groups, with 10 samples in each.

Group 1 made use of the innovative light GBM

classifier, whereas Group 2 adopted the Adaboost

classifier. For the desired accuracy, samples were

sourced from the device and underwent ten

repetitions with an 80% G power, a significance level

of 0.05%, and a 95% confidence interval

(Kakaraparthi & Karthick, 2022). The dataset, a

compilation of Sepsis Clinical Data from patients,

was accessed via the Physionet Repository.

The Jupyter Notebook served as the coding

platform for testing, whilst SPSS version 26.0.1 was

deployed for statistical data analyses. The operations

ran on a laptop equipped with an Intel Core i5

processor and a 16GB RAM. The Physionet

repository supplied the dataset (Reyna et al., 2019).

Out of the acquired training set, 20,336 psv files,

encompassing patient clinical data in pipe-delimited

text formats, were subsequently converted to CSV.

The dataset exhibited an imbalance, with nearly 60%

of its data being null values. Resampling techniques

were used to balance the dataset. Initially containing

41 columns, post feature engineering, it was trimmed

down to 14 columns. These columns represented vital

signs, lab results, demographics, and outcomes.

Among the 14, 13 were independent whilst one was

dependent. The 'Sepsis Label' column indicates

whether a patient has sepsis.

Innovative Novel LightGBM Classifier

The innovative LightGBM classifier stands as a

machine learning model tailored for binary or

multiclass classification assignments. It's a gradient-

boosting platform known for its high performance,

which employs tree-based learning algorithms to craft

a decision tree ensemble. Owing to its design,

LightGBM is adept at efficiently handling large

datasets and real-time tasks. The model incorporates

a distinctive technique named "Gradient-based One-

Side Sampling" (GOSS) to refine the gradient

boosting procedure. This not only slashes memory

consumption but also accelerates the training phase.

Additionally, LightGBM is equipped with advanced

functionalities such as managing categorical features,

early termination, and bespoke loss functions,

rendering it an invaluable asset for classification

undertakings that demand speed and precision

(Hecht-Nielsen, 2020).

Algorithm 1.

Input: The clinical dataset of patients.

Output: Predicted label of sepsis.

Step 1: The necessary packages are imported.

Step 2: Load the dataset and store it as a data frame

using pandas.

Step 3: Calculate the null values and eliminate

them using imputation.

Step 4: Evaluate the component significance and

extract the important features using XGBoost.

Step 5: Make the labels normalized by using Label

Encoder.

Step 6: Create the training and testing datasets

using Sklearn libraries.

Step 7: Find the parameter combinations by

performing the hyper parameter tuning Boost

Machine Classifier.

Step 8: Using the parameters found by hyper

parameter tuning, Initialise the Light Gradient

boosting machine classifier.

Step 9: Commence the training of the Light

Gradient Boost Machine Classifier utilizing the

provided training data.

Step 10: The performance of the model is evaluated

by validating it with the provided testing data.

Step 11: Using Sklearn metrics compute the

accuracy and plot using matplotlib

Adaboost Classifier

AdaBoost, short for Adaptive Boosting, is a

supervised machine learning technique suitable for

both classification and regression tasks. Acting as a

meta-algorithm, it amalgamates the predictions from

numerous weaker classifiers to forge a robust and

more precise classifier. AdaBoost functions

iteratively, sequentially training weaker classifiers,

with each new one focusing on amending the errors

of the preceding one. A pivotal attribute of AdaBoost

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

596

is its proficiency in managing imbalanced datasets,

wherein certain classes might be underrepresented

compared to others. The flexibility to integrate with a

plethora of base classifiers bolsters its versatility,

rendering it a favoured choice in the realm of machine

learning. On the whole, AdaBoost is celebrated for its

prowess to augment classification model accuracy

and provide considerable interpretability across

diverse domains. (Hao & Huang, 2023).

Algorithm 2.

Input: The clinical dataset of patients.

Output: Predicted label of sepsis.

Step 1: The required packages are imported.

Step 2: Load the dataset using pandas and store it

in a dataframe.

Step 3: The dataset is checked to ensure whether it

is balanced or not and perform resampling.

Step 4: Find the correlation between labels and

compute component significance.

Step 5: Extract the most important features using

XGBoost.

Step 6: Normalize the dataset using label encoder.

Step 7: Use the hyper parameter tuning to identify

the best optimal parameter combinations.

Step 8: Using the parameters found in the last step,

initialize the Adaboost classifier.

Step 9: Train the Adaboost classifier with training

data.

Step 10: Determine the Accuracy and Log loss for

test data.

Statistical Analysis

The statistical examination of the suggested and

counterpart algorithms was conducted utilising the

IBM SPSS 26.0.1 software. In the clinical dataset,

'Sepsis Label' serves as the dependent variable,

whereas the independent variables encompass HR,

O2SAT, Temp, SBP, MAP, DBP, RESP, EtC02,

BaseExcess, HCO3, FiO2, pH, PaCO2, SaO2, AST,

BUN, Alkalinephos, Calcium, Chloride, Creatinine,

Bilirubin_direct, Glucose, Lactate, Magnesium,

Phosphate, Potassium, Bilirubin_total, Troponin,

Hct, Hgb, PTT, WBC, Fibrinogen, Platelets, Age,

Gender, Unit1, Unit2, HospAdmTime, and ICULOS.

An independent sample T-test was employed for both

the proposed and the contrasting algorithms. Post

analysis, metrics like mean accuracy, standard

deviation, and standard error were documented

(Hussain et al., 2022).

3 RESULTS

In this research study, two algorithms – the

Innovative Novel LightGBM Technique and the

AdaBoost Classifier – were utilised, with accuracy as

the primary performance metric. The AdaBoost

Classifier's performance rendered an accuracy of

77.58%, which is comparatively lower than that of the

Innovative Novel LightGBM Technique, which

achieved an impressive accuracy of 96.41%.

Table 1: The precise scores of both Light GBM and

Adaboost Classifier models, based on a sample size of 10

each, are presented. The LGBM classifier model exhibits

accuracies ranging from 98.63% to 93.45%, while the

Adaboost classifier model displays accuracies ranging from

79.82% to 76.43%.

S. No LGBM Adaboost

1 93.45 74.58

2 94.56 75.67

3 95.23 76.43

4 96.46 76.81

5 96.45 77.23

6 96.83 77.77

7 97.21 78.63

8 97.61 78.92

9 98.63 78.94

10 98.76 79.82

Table 1 enumerates the accuracy rates acquired

across 10 iterations for both Group 1 and Group 2.

Meanwhile, Table 2 highlights the mean accuracies,

standard deviation, and standard error mean derived

from group statistics. The LightGBM model

registered a mean accuracy of 96.41%, while the

AdaBoost classifier marked a mean accuracy of

77.58%. Table 3 showcases the results of the

Independent Samples T-test performed in SPSS,

revealing a significance value of p=0.019 (p<0.05).

This signifies a statistical difference between the two

groups under study. Figure 1 offers a bar graph

juxtaposing the Innovative Novel LightGBM

Technique and the AdaBoost classifier, plotting the

variables of mean accuracy and loss on the Y-axis.

The Innovative LightGBM Technique exhibits

superior relevance compared to its AdaBoost

counterpart. Furthermore, the error bars within the

graph facilitate an assessment of the error rate,

highlighting that the Innovative LightGBM

Technique possesses a notably reduced error rate in

contrast to the AdaBoost Classifier.

Prediction of Sepsis Using Light Gradient-Boosting Machine Classiﬁer in Comparison with Adaboost Classiﬁer Based on Accuracy

597

Table 2: The Innovative LightGBM Classifier has a mean accuracy of 96.41%, while the Adaboost Classifier's is 77.58%.

Standard deviation and standard error were calculated for both groups. LightGBM showed a higher standard deviation than

Adaboost.

Algorith

N Mean Standard Deviation Standard Error Mean

LGBM

ADABOOST

96.41

77.58

1.73804

1.7

0.54962

0.52429

Table 3: The Independent Samples T-test shows that the p-value is p=0.019 (p<0.05), which indicates that there is a significant

difference between the two groups. The mean accuracy of the two groups was compared assuming equal variances, and a

95% confidence interval was used.

Levene’s test

for equality o

variances

T-test for Equality of Means

F Sig. t df

Sig. 2-

tailed

Mean

Difference

Std. Error

Difference

95% confidence interval

of the difference

Lowe

Uppe

Accuracy

Equal

variance

Assumed

0.010 0.921 24.148 18 0.019 18.83900 0.78013 17.200000 20.47800

Accuracy

Equal

variance not

Assume

24.148 17.999 0.019 18.83900 0.78013 17.200000 20.47800

Figure 1: A graphical comparison of LGBM and Adaboost Classifiers based on mean accuracy and loss. LGBM outperforms

with higher accuracy and lower loss. Both classifiers are plotted on the x-axis against accuracy and loss on the y-axis, with a

95% confidence interval of +/-2 SD.

4 DISCUSSION

The findings of this research study highlight the

superior performance of the Innovative LightGBM

Technique over the AdaBoost Classifier in predicting

sepsis. The significance value, calculated using the

independent sample T-Test, stood at 0.019 (p<0.05),

marking the research as statistically significant. The

Innovative LightGBM Technique recorded a

commendable accuracy rate of 96.41% coupled with

a log loss of 0.064, clearly outshining the AdaBoost

Classifier which posted an accuracy of 77.58% and a

loss of 0.6731.

This study resonates with the contention that a

lower rate of loss is invaluable as it attests to the

efficiency of the approach (Elith et al., 2008). In their

discussion on the innovative technique known as

boosted regression trees, researchers pointed out the

application of boosting algorithms like AdaBoost for

challenges like two-class classification. This model

accentuated the significance of applying weights to

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

598

observations, underscoring those that are weakly

modelled. Several published research articles

corroborate these findings. For instance, Nesaragi &

Patidar (2021) conceptualised an ensemble model

amalgamating LightGBM, XGBoost, and Random

Forest, marking their best performance with an AUC

of 0.792 and an ACC of 0.727. Meanwhile, Bhavekar

& Goswami (2022) employed the five-fold-cross-

validation method to achieve an average normalized

utility score of 0.4314. L. Peng et al. (2022)

constructed seven diverse models, with the light

GBM model emerging as the best, recording an

accuracy of 0.96 on the test dataset. Chami &

Tavakolian (2019) benchmarked the Light Gradient

Boosting Machine Classifier against a hybrid of

survival analysis and neural networks, demonstrating

the supremacy of LGBM with a score of 0.172.

However, contrary perspectives are also evident.

Tarif et al. (2018) and Neelagandan (2012) observed

superior accuracy and efficiency with AdaBoost

classifiers, especially when juxtaposed with gradient-

boosted tree classifiers.

Nonetheless, this study isn't without limitations.

While the Innovative LightGBM Technique offers

impressive performance, it can be more time-

consuming during training phases and can consume

more memory compared to other classifiers. This

becomes more evident with extensive datasets or

when using specific high-category categorical

variables. There's also a susceptibility to overfitting,

especially with noisy data or over-extended training

durations. Looking ahead, the aspiration is to refine

the research by incorporating deep learning models.

Despite their promise of potentially unparalleled

accuracy in sepsis prediction, these models demand

extended training durations and necessitate advanced

computational infrastructure. Enhancing accuracy is

pivotal, for it directly impacts the mortality rate,

thereby ensuring better patient outcomes.

5 CONCLUSION

Drawing from the extensive analysis and findings of

this research, it becomes unequivocal that machine

learning models, especially the ones tailored for

specialized tasks, have the potential to revolutionize

the medical diagnostics sector. The comparative

analysis between the Innovative LightGBM

Technique and the AdaBoost classifier in the context

of sepsis prediction is a testament to this. Based on

our comprehensive discussions, the following six key

points emerge:

Performance Metrics: The Innovative LightGBM

Technique, with an impressive accuracy of 96.41%,

starkly outperformed the AdaBoost classifier, which

only managed to secure an accuracy of 77.58%.

Accuracy being a critical metric in medical diagnosis,

this difference in performance can translate to

tangible improvements in patient care.

Handling of Large Datasets: LightGBM is known

for its efficiency and scalability, which makes it adept

at handling large datasets. The ability to effectively

deal with extensive data is critical in medical

applications where vast amounts of patient data are

often involved.

Boosting Techniques: LightGBM employs

advanced boosting techniques such as Gradient-based

One-Side Sampling (GOSS). This not only

accelerates the training process but also optimizes

memory usage, making the model both fast and

resource-efficient.

Mitigation of Overfitting: Overfitting is a

perennial concern in machine learning, more so in

medical diagnostics. While the LightGBM model did

show potential susceptibilities to overfitting,

especially with noisy data, its performance in this

research still overshadowed the AdaBoost Classifier.

Versatility of AdaBoost: Despite the lower

accuracy, it's important to recognize the versatility of

the AdaBoost classifier. Its iterative approach to

rectifying errors and its compatibility with a range of

base classifiers still make it a valuable tool in many

applications.

Future Direction: While the LightGBM model has

shown superior performance in this research, it also

brings forth the idea of exploring deep learning

models in the future. The aim would be to achieve

even higher accuracy levels, albeit with the

understanding that these models might require more

intensive computational resources.

In conclusion, the overarching insight is that the

Innovative LightGBM Technique provides a more

accurate and efficient means of predicting sepsis

compared to the AdaBoost classifier. This not only

has implications for the advancement of machine

learning in healthcare diagnostics but also

underscores the critical role of selecting the

appropriate model for specific challenges.

REFERENCES

Bhavekar, G. S., & Goswami, A. D. (2022). Herding

Exploring Algorithm With Light Gradient Boosting

Machine Classifier for Effective Prediction of Heart

Diseases. In International Journal of Swarm

Prediction of Sepsis Using Light Gradient-Boosting Machine Classiﬁer in Comparison with Adaboost Classiﬁer Based on Accuracy

599

Intelligence Research (Vol. 13, Issue 1, pp. 1–22).

https://doi.org/10.4018/ijsir.302609

Bloch, E., Rotem, T., Cohen, J., Singer, P., & Aperstein, Y.

(2019). Machine Learning Models for Analysis of Vital

Signs Dynamics: A Case for Sepsis Onset Prediction.

Journal of Healthcare Engineering, 2019.

https://doi.org/10.1155/2019/5930379

Chami, S., & Tavakolian, K. (2019). Comparative Study of

Light-GBM and LSTM for Early Prediction of Sepsis

From Clinical Data. In 2019 Computing in Cardiology

Conference (CinC).

https://doi.org/10.22489/cinc.2019. 367

Deepak., John Justin Thangaraj, S., & Rajesh Khanna, M.

(2020, October 7). An improved early detection method

of autism spectrum anarchy using euclidean method.

2020 Fourth International Conference on I-SMAC (IoT

in Social, Mobile, Analytics and Cloud) (I-SMAC).

2020 Fourth International Conference on I-SMAC (IoT

in Social, Mobile, Analytics and Cloud) (I-SMAC),

Palladam, India. https://doi.org/10.1109/i-smac49090.

2020.9243361

Elith, J., Leathwick, J. R., & Hastie, T. (2008). A working

guide to boosted regression trees. The Journal of

Animal Ecology, 77(4), 802–813.

G. Ramkumar, R. Thandaiah Prabu, Ngangbam Phalguni

Singh, U. Maheswaran, Experimental analysis of brain

tumor detection system using Machine learning

approach, Materials Today: Proceedings, 2021, ISSN

2214-7853,

https://doi.org/10.1016/j.matpr.2021.01.246.

Hao, L., & Huang, G. (2023). An improved AdaBoost

algorithm for identification of lung cancer based on

electronic nose. Heliyon, 9(3), e13633.

Hecht-Nielsen, R. (2020). LPG-model: A novel model for

throughput prediction in stream processing, using a

light gradient boosting machine, incremental principal

component analysis, and deep gated recurrent unit

network. Information Sciences, 535, 107–129.

Hussain, M. M., Mahammad Hussain, M., & Karthick, V.

(2022). Efficient Search in Cloud Storage with Reduced

Computational Cost using Token Generation Method

over Crypto Hash Algorithm. In 2022 3rd International

Conference on Smart Electronics and Communication

(ICOSEC). https://doi.org/10.1109/icosec54921.2022.

9952137

Kakaraparthi, A., & Karthick, V. (2022). A Secure and

Cost-Effective Platform for Employee Management

System Using Lightweight Standalone Framework

Over Diffie Hellman’s Key Exchange Algorithm. In

ECS Transactions (Vol. 107, Issue 1, pp. 13663–

13674). https://doi.org/10.1149/10701.13663ecst

Liu, R., Greenstein, J. L., Granite, S. J., Fackler, J. C.,

Bembea, M. M., Sarma, S. V., & Winslow, R. L.

(2019). Data-driven discovery of a novel sepsis pre-

shock state predicts impending septic shock in the ICU.

Scientific Reports, 9(1), 1–9.

Neelagandan, R. (2012). High-Performance Face

Detection Using McT and Adaboost Algorithm. LAP

Lambert Academic Publishing.

Nesaragi, N., & Patidar, S. (2021). An Explainable Machine

Learning Model for Early Prediction of Sepsis Using

ICU Data. In Infections and Sepsis Development.

https://doi.org/10.5772/intechopen.98957

Peng, L., Peng, C., Yang, F., Wang, J., Zuo, W., Cheng, C.,

Mao, Z., Jin, Z., & Li, W. (2022). Machine learning

approach for the prediction of 30-day mortality in

patients with sepsis-associated encephalopathy. BMC

Medical Research Methodology, 22(1), 183.

Peng, X., Ding, Y., Wihl, D., Gottesman, O., Komorowski,

M., Lehman, L.-W. H., Ross, A., Faisal, A., & Doshi-

Velez, F. (2018). Improving Sepsis Treatment

Strategies by Combining Deep and Kernel-Based

Reinforcement Learning. AMIA ... Annual Symposium

Proceedings / AMIA Symposium. AMIA Symposium,

2018, 887–896.

Pravda, J. (2021). Sepsis: Evidence-based pathogenesis and

treatment. Pediatric Critical Care Medicine: A Journal

of the Society of Critical Care Medicine and the World

Federation of Pediatric Intensive and Critical Care

Societies, 10(4), 66.

Reyna, M., Josef, C., Jeter, R., Shashikumar, S., Moody, B.,

Brandon Westover, M., Sharma, A., Nemati, S., &

Clifford, G. D. (2019). Early Prediction of Sepsis from

Clinical Data: The PhysioNet/Computing in

Cardiology Challenge 2019 [Data set].

https://doi.org/10.13026/ v64v-d857

Sivakumar, V. L., Nallanathel, M., Ramalakshmi, M., &

Golla, V. (2022). Optimal route selection for the

transmission of natural gas through pipelines in

Tiruchengode Taluk using GIS–a preliminary study.

Materials Today: Proceedings, 50, 576-581.

Shrestha, U., Alsadoon, A., Prasad, P. W. C., Al Aloussi,

S., & Alsadoon, O. H. (2021). Supervised machine

learning for early predicting the sepsis patient: modified

mean imputation and modified chi-square feature

selection. Multimedia Tools and Applications, 80(13),

20477–20500.

Tarif, A. M., Raju, S. M., Al Amin Ashik, M., Islam, M. S.,

& Tahera, T. (2018). Self-Driving Car Simulation using

Adaboost-CNN Algorithm. GRIN Verlag.

Taylor, R. A., Pare, J. R., Venkatesh, A. K., Mowafi, H.,

Melnick, E. R., Fleischman, W., & Hall, M. K. (2016).

Prediction of In-hospital Mortality in Emergency

Department Patients With Sepsis: A Local Big Data-

Driven, Machine Learning Approach. Academic

Emergency Medicine: Official Journal of the Society

for Academic Emergency Medicine, 23(3), 269–278.

Wong, H. R., Cvijanovich, N. Z., Anas, N., Allen, G. L.,

Thomas, N. J., Bigham, M. T., Weiss, S. L., Fitzgerald,

J., Checchia, P. A., Meyer, K., Shanley, T. P., Quasney,

M., Hall, M., Gedeit, R., Freishtat, R. J., Nowak, J.,

Shekhar, R. S., Gertz, S., Dawson, E., … Lindsell, C. J.

(2015). Developing a Clinically Feasible Personalized

Medicine Approach to Pediatric Septic Shock.

American Journal of Respiratory and Critical Care

Medicine. https://doi.org/10.1164/rccm.201410-

1864OC

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

600