Prediction of gestational diabetes using deep learning and Bayesian optimization and traditional machine learning techniques

Kurt, Burçin; Gürlek, Beril; Keskın, Seda; Özdemir, Sinem; Karadeniz, Özlem; Kırkbir, İlknur Buçan; Kurt, Tuğba; Ünsal, Serbülent; Kart, Cavit; Baki, Neslihan; Turhan, Kemal

doi:10.1007/s11517-023-02800-7

Cited by 5 publications

(1 citation statement)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the field of classification and diagnostics, there have been developments such as the BO-SVM model for Parkinson's disease [9] and an automation system that utilizes CNN and KNN for the diagnosis of Alzheimer's [10]. In the field of prediction, Bayesian optimization has been used for diverse applications including predicting jump height after the release of ice on the transmission [11], cryptocurrency price prediction [12], forecasting the number of vehicles on the road [13], predicting the collapse of transmission foundation towers [14], real-time prediction of electric load on a smart grid [15], prediction of wind energy generation [16], and a decision support prediction system for the diagnosis of gestational diabetes [17]. Furthermore, Bayesian optimization has also been employed in predicting the state of health (SOH) of lithium-ion batteries [18].…”

Section: Introductionmentioning

confidence: 99%

Facebook Prophet Model with Bayesian Optimization for USD Index Prediction

Hamdani,

Swanjaya,

Helilintar

2023

JUITA

View full text Add to dashboard Cite

Accuracy is the primary focus in prediction research. Optimization is conducted to improve the performance of prediction models, thereby enhancing prediction accuracy. This study aims to optimize the Facebook Prophet model by performing hyperparameter tuning using Bayesian Optimization to improve the accuracy of USD Index Value prediction. Evaluation is conducted through multiple prediction experiments using different ranges of historical data. The results of the study demonstrate that performing hyperparameter tuning on the Facebook Prophet model yields better prediction results. Prior to parameter tuning, the MAPE indicator metric is 1.38% for the historical data range of 2014-2023, and it decreases to 1.33% after parameter tuning. Further evaluation shows improved prediction performance using different ranges of historical data. For the historical data range of 2015-2023, the MAPE value decreases from 1.39% to 1.20%. Similarly, for the data range of 2016-2023, the MAPE decreases from 1.12% to 0.80%. Furthermore, for the data range of 2017-2023, there is a decrease from 0.80% to 0.76%. This is followed by the data range of 2018-2023, with a decrease from 0.75% to 0.70%. Lastly, for the data range of 2019-2023, there is a decrease from 0.63% to 0.55%. These results demonstrate that performing Hyperparameter Optimization using Bayesian Optimization consistently improves prediction accuracy in the Facebook Prophet model.

show abstract

Section: Introductionmentioning

confidence: 99%

Facebook Prophet Model with Bayesian Optimization for USD Index Prediction

Hamdani,

Swanjaya,

Helilintar

2023

JUITA

View full text Add to dashboard Cite

show abstract

A novel intelligent model for visualized inference of medical diagnosis: A case of TCM

Qi-yu,

Wen-heng,

Jia-fen

et al. 2024

Artificial Intelligence in Medicine

View full text Add to dashboard Cite

Hi-Le and HiTCLe: Ensemble Learning Approaches for Early Diabetes Detection Using Deep Learning and Explainable Artificial Intelligence

Shaheen,

Javaid,

Alrajeh

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Diabetes is a metabolic disease caused by the body's failure to use insulin or break down meals correctly. Every year, an alarming number of new cases of diabetes are recorded. A poor lifestyle and an unfavorable environment are the two main causes of diabetes. If it is not treated at early stages, it becomes a lifelong disease and further leads to failure of important organs such as the kidneys, heart, eyes, and so on. This danger can be decreased with timely and precise identification. Deep Learning (DL) is the best method for illness prediction, as demonstrated by recent developments in DL for clinical use. We have proposed two ensemble learning approaches: blending and hybrid by using the Diabetes Prediction Dataset (DPD), which is a highly imbalanced dataset. The number of diabetic patients in it are 8500 whereas, the number of non-diabetic individuals are 91500. To overcome the class imbalance problem, a Proximity-Weighted Synthetic Oversampling (ProWSyn) technique is implemented. We have proposed a hybrid of highway and LeNet model, named Hi-Le, for early and accurate diabetes detection. Hi-Le model achieves an accuracy of 94%, a F1-Score of 96%, precision score of 94% and recall of 95% and beats its individual models in terms of accuracy, F1-Score, precision and recall. We have also proposed a blending model named HiTCLe using Highway, LeNet, and a Temporal Convolutional Network (TCN) to detect and predict diabetes at an early stage. HiTCLe performs best, beats its individual models, highway, TCN and LeNet, and achieves an accuracy score of 94% and a F1-Score of 94%, whereas individual models achieve an accuracy score between 89% and 91% on 10 epochs. To validate models' results, we have implemented K-Fold Cross Validation (K-FCV). Also, to know the features contributions, we have implemented Shapley Additive eXplanations (SHAP) post processing technique. Both ensemble learning models outperform their individual models in term of accurate diabetes detection and prediction.

show abstract

Prediction of gestational diabetes using deep learning and Bayesian optimization and traditional machine learning techniques

Cited by 5 publications

References 28 publications

Facebook Prophet Model with Bayesian Optimization for USD Index Prediction

Facebook Prophet Model with Bayesian Optimization for USD Index Prediction

A novel intelligent model for visualized inference of medical diagnosis: A case of TCM

Hi-Le and HiTCLe: Ensemble Learning Approaches for Early Diabetes Detection Using Deep Learning and Explainable Artificial Intelligence

Contact Info

Product

Resources

About