Diabetes Prediction Model Research Articles

Diabetes Risk Prediction Using Extreme Gradient Boosting (XGBoost)

One of the uses of medical data from diabetes patients is to produce models that can be used by medical personnel to predict and identify diabetes in patients. Various techniques are used to be able to provide a diabetes model as early as possible based on the symptoms experienced by diabetic patients, including using machine learning. The machine learning technique used to predict diabetes in this study is extreme gradient boosting (XGBoost). XGBoost is an advanced implementation of gradient boosting along with multiple regularization factors to accurately predict target variables by combining simpler and weaker model set estimations. Errors made by the previous model are tried to be corrected by the next model by adding some weight to the model. The diabetes prediction model using XGBoost is shown in the form of a tree, with the accuracy of the model produced in this study of 98.71%

Diabetes prediction model using data mining techniques

Diabetes is the leading cause of death in the world, and it also affects kidney disease, loss of vision, and heart disease. Data mining techniques contribute to health care decisions for accurate disease diagnosis and treatment, reducing the workload of experts. Diabetes prediction is a rapidly expanding field of research. Early diabetes prediction will result in improved treatment. Diabetes causes a variety of health issues. Therefore, it is critical to prevent, monitor, and raise awareness about it. Type 1 and Type 2 diabetes can cause heart disease, renal problems, and eye difficulties. In this paper, we propose a diabetes prediction model using data mining techniques. We apply four data mining techniques such as Random Forest, Support Vector Machine (SVM), Logistic Regression, and Naive Bayes. The proposed mechanism is trained using Python and analysed with a real dataset, which is collected from Kaggle. Furthermore, the performance of the proposed mechanism is analysed using the confusion matrix, sensitivity and accuracy performance metrices. In logistic regression, the accuracy is high, i.e., 82.46%, in comparison to other data mining techniques.

Germline Genetic and Treatment-Related Risk Factors for Diabetes Mellitus in Survivors of Childhood Cancer: A Report From the Childhood Cancer Survivor Study and St Jude Lifetime Cohorts.

To characterize germline genetic risk factors of diabetes mellitus among long-term survivors of childhood cancer. Adult survivors of childhood cancer from the Childhood Cancer Survivor Study (CCSS) Original Cohort (n = 5,083; 383 with diabetes) were used to conduct a discovery genome-wide association study. Replication was performed using the CCSS Expansion (n = 2,588; 40 with diabetes) and the St Jude Lifetime (SJLIFE; n = 3,351; 208 with diabetes) cohorts. Risk prediction models, stratified on exposure to abdominal radiation, were calculated using logistic regression including attained age, sex and body mass index, diagnosis, alkylating chemotherapy, age at cancer diagnosis, and a polygenic risk score (PRS) on the basis of 395 diabetes variants from the general population. Area under the receiver operating characteristic curve (AUC) was calculated for models on the basis of traditional risk factors, clinical risk factors, and PRS. There was a genome-wide significant association of rs55849673-A with diabetes among survivors (odds ratio, 2.9; 95% CI, 2.0 to 4.2; P = 3.7 × 10-8), which is related to expression of ERCC6L2 in the Genotype-Tissue Expression project. The association of rs55849673-A was observed largely among survivors not exposed to abdominal radiation (odds ratio = 3.5, P = 1.1 × 10-7) and the frequency of rs55849673-A was consistently higher among diabetic survivors in the CCSS Expansion and SJLIFE cohorts. Risk prediction models including traditional diabetes risk factors, clinical risk factors and PRS had an optimism-corrected AUC of 0.801, with an AUC of 0.751 in survivors treated with abdominal radiation versus 0.813 in survivors who did not receive abdominal radiation. There is evidence for a novel locus of diabetes among survivors not exposed to abdominal radiation. Further refinement and validation of clinic-based risk prediction models for diabetes among long-term survivors of childhood cancer is warranted.

Improving Machine Learning Diabetes Prediction Models for the Utmost Clinical Effectiveness.

The early prediction of diabetes can facilitate interventions to prevent or delay it. This study proposes a diabetes prediction model based on machine learning (ML) to encourage individuals at risk of diabetes to employ healthy interventions. A total of 38,379 subjects were included. We trained the model on 80% of the subjects and verified its predictive performance on the remaining 20%. Furthermore, the performances of several algorithms were compared, including logistic regression, decision tree, random forest, eXtreme Gradient Boosting (XGBoost), Cox regression, and XGBoost Survival Embedding (XGBSE). The area under the receiver operating characteristic curve (AUROC) of the XGBoost model was the largest, followed by those of the decision tree, logistic regression, and random forest models. For the survival analysis, XGBSE yielded an AUROC exceeding 0.9 for the 2- to 9-year predictions and a C-index of 0.934, while the Cox regression achieved a C-index of 0.921. After lowering the threshold from 0.5 to 0.25, the sensitivity increased from 0.011 to 0.236 for the 2-year prediction model and from 0.607 to 0.994 for the 9-year prediction model, while the specificity showed negligible changes. We developed a high-performance diabetes prediction model that applied the XGBSE algorithm with threshold adjustment. We plan to use this prediction model in real clinical practice for diabetes prevention after simplifying and validating it externally.

Polygenic scores for low lung function and the future risk of adverse health outcomes

AimsReduced lung function and adverse health outcomes are often observed. This study characterizes genetic susceptibility for reduced lung function and risk of developing a range of adverse health outcomes.MethodsWe studied 27,438 middle-aged adults from the Malmö Diet and Cancer study (MDCS), followed up to 28.8 years. Trait-specific Polygenic scores (PGS) for forced expiratory volume in 1 s (FEV1) and forced vital capacity (FVC) were constructed for each participant using MDCS genetic data and summary statistics from the latest GWAS of lung function. Linear regression models and cox proportional hazards regression models were used to assess associations between adverse health outcomes and lung function-PGS.ResultsFEV1-PGS and FVC-PGS were significantly associated with mean sBP at baseline after adjustments (FEV1-PGS Q1 (highest PGS = highest lung function): 140.7mmHg vs. Q4: 141.5mmHg, p-value 0.008). A low FVC-PGS was significantly associated with the risk of future diabetic events after adjustments (Q4 vs. Q1 HR: 1.22 (CI 1.12–1.32), p-trend < 0.001) and had added value to risk prediction models for diabetes. Low FEV1-PGS was significantly associated with future coronary events (Q4 vs. Q1 HR: 1.13 (CI: 1.04–1.22), p-trend 0.008). No significant association was found between PGS and sudden cardiac death, chronic kidney disease or all-cause mortality. Results remained largely unchanged in a subgroup of subjects when further adjusted for apolipoproteins.ConclusionGenetic susceptibility for reduced lung function is associated with higher sBP, increased risk of diabetes and to a lesser extent, future coronary events, suggesting etiological roles of lung function on these outcomes. Using PGS, high-risk groups could be early detected to implement early lifestyle changes to mitigate the risk.

Prognostic factors in diabetes: Comparison of Chi-square automatic interaction detector (CHAID) decision tree technology and logistic regression.

This study aimed to develop a diabetes prediction model. The model performance was compared with logistic regression, and the decision tree Chi-square automatic interaction detection (CHAID) was used to predict diabetes.In total, 3233 patients were included in the analysis. Of these, 589 patients with diabetes and 2644 patients without diabetes were included after analyzing the study sample from the Korean Genome and Epidemiology Study (KoGES)-8 data. In this study, Diabetes Mellitus (DM) diagnosis prediction was compared with logistic regression and prediction through machine learning (ML) using the CHAID decision classification tree. We performed statistical analysis using the CHAID method with International Business Machine (IBM) statistical program SPSS®.We performed logistic regression analysis to predict the classification of diabetes accurately, and the total classification accuracy of the analysis was 81.7%, and the CHAID decision tree classification accuracy was 81.8%. A diabetes diagnosis decision tree was created, which included seven terminal nodes and three depth levels. This analysis showed that a blood pressure problem and hospital visits were the most decisive variables at the time of classification, and two risk levels were created for diabetes diagnosis.The suggested method is a valuable tool for predicting diabetes. Patients who visit the hospital because of blood pressure problems are more likely to develop diabetes than under-treating hyperlipidemia. The diabetes prediction model can help doctors make decisions by detecting the possibility of diabetes early; however, it is impossible to diagnose diabetes using only the model without the doctor’s opinion.

Effective Approach for Early Detection of Diabetes by Logistic Regression through Risk Prediction

Heart disease, cancer, renal failure, eye damage, and blindness are just some of the complications that may result from uncontrolled diabetes. Scientists are inspired to develop a Machine Learning (ML) approach for diabetes forecasting. To improve illness diagnosis, medical personnel must make use of ML algorithms. Different ML algorithms for identifying diabetes risk at an early stage are examined and contrasted in this research. The goal in analysing diabetes prediction models is to develop criteria for selecting high-quality studies and synthesising the results from several studies. Nonlinearity, normality, correlation structure, and complexity characterise the vast majority of medical data, making analysis of diabetic data a formidable task. Algorithms based on machine learning are not permitted to be used in healthcare or medical imaging. Early diabetes mellitus prediction necessitates a strategy distinct from those often used. Diabetic patients and healthy individuals may be separated using a risk stratification approach based on machine learning. This study is highly recommended since it reviews a variety of papers that may be used by researchers working on diabetes prediction models.

Gradient boosting decision tree becomes more reliable than logistic regression in predicting probability for diabetes with big data

We sought to verify the reliability of machine learning (ML) in developing diabetes prediction models by utilizing big data. To this end, we compared the reliability of gradient boosting decision tree (GBDT) and logistic regression (LR) models using data obtained from the Kokuho-database of the Osaka prefecture, Japan. To develop the models, we focused on 16 predictors from health checkup data from April 2013 to December 2014. A total of 277,651 eligible participants were studied. The prediction models were developed using a light gradient boosting machine (LightGBM), which is an effective GBDT implementation algorithm, and LR. Their reliabilities were measured based on expected calibration error (ECE), negative log-likelihood (Logloss), and reliability diagrams. Similarly, their classification accuracies were measured in the area under the curve (AUC). We further analyzed their reliabilities while changing the sample size for training. Among the 277,651 participants, 15,900 (7978 males and 7922 females) were newly diagnosed with diabetes within 3 years. LightGBM (LR) achieved an ECE of 0.0018 ± 0.00033 (0.0048 ± 0.00058), a Logloss of 0.167 ± 0.00062 (0.172 ± 0.00090), and an AUC of 0.844 ± 0.0025 (0.826 ± 0.0035). From sample size analysis, the reliability of LightGBM became higher than LR when the sample size increased more than 10^4. Thus, we confirmed that GBDT provides a more reliable model than that of LR in the development of diabetes prediction models using big data. ML could potentially produce a highly reliable diabetes prediction model, a helpful tool for improving lifestyle and preventing diabetes.

Predictive Analysis of Diabetes-Risk with Class Imbalance.

Diabetes type 2 (T2DM) is a common chronic disease, increasingly leading to many complications and affecting vital organs. Hyperglycemia is the main characteristic caused by insufficient insulin secretion and poses a serious risk to human health. The objective is to construct a type-2 diabetes prediction model with high classification accuracy. Advanced machine learning and predictive model techniques are utilized to achieve cutting-edge techniques for the early diagnosis of diabetes. This paper proposes an efficient performance model to predict and classify the minority class of type-2 diabetes. The impact of oversampling and undersampling approaches to reduce the effect of an unbalanced class has been compared to classification performance algorithms. Synthetic Minority Oversampling (SMOTE) and Tomek-links techniques are applied and examined. The outcomes were then compared to the original unbalanced dataset using an artificial neural network (ANN) predictive model. The model is compared with other state-of-the-art classifiers such as support vector machine (SVM), random forest (RF), and decision tree (DT). The tuned model had the best accuracy of 92.2%. The experimental findings clearly manifest the improvement in accuracy and evaluation metrics in terms of AUC and F1-measure using the SMOTE oversampling strategy rather than the baseline and undersampling schemes. The study recommends adopting dynamic hyperparameter optimization to further improve accuracy.

Quantifying the utility of islet autoantibody levels in the prediction of type 1 diabetes in children

Aims/hypothesisThe aim of this study was to explore the utility of islet autoantibody (IAb) levels for the prediction of type 1 diabetes in autoantibody-positive children.MethodsProspective cohort studies in Finland, Germany, Sweden and the USA followed 24,662 children at increased genetic or familial risk of developing islet autoimmunity and diabetes. For the 1403 who developed IAbs (523 of whom developed diabetes), levels of autoantibodies against insulin (IAA), glutamic acid decarboxylase (GADA) and insulinoma-associated antigen-2 (IA-2A) were harmonised for analysis. Diabetes prediction models using multivariate logistic regression with inverse probability censored weighting (IPCW) were trained using 10-fold cross-validation. Discriminative power for disease was estimated using the IPCW concordance index (C index) with 95% CI estimated via bootstrap.ResultsA baseline model with covariates for data source, sex, diabetes family history, HLA risk group and age at seroconversion with a 10-year follow-up period yielded a C index of 0.61 (95% CI 0.58, 0.63). The performance improved after adding the IAb positivity status for IAA, GADA and IA-2A at seroconversion: C index 0.72 (95% CI 0.71, 0.74). Using the IAb levels instead of positivity indicators resulted in even better performance: C index 0.76 (95% CI 0.74, 0.77). The predictive power was maintained when using the IAb levels alone: C index 0.76 (95% CI 0.75, 0.76). The prediction was better for shorter follow-up periods, with a C index of 0.82 (95% CI 0.81, 0.83) at 2 years, and remained reasonable for longer follow-up periods, with a C index of 0.76 (95% CI 0.75, 0.76) at 11 years. Inclusion of the results of a third IAb test added to the predictive power, and a suitable interval between seroconversion and the third test was approximately 1.5 years, with a C index of 0.78 (95% CI 0.77, 0.78) at 10 years follow-up.Conclusions/interpretationConsideration of quantitative patterns of IAb levels improved the predictive power for type 1 diabetes in IAb-positive children beyond qualitative IAb positivity status.Graphical abstract

An Ensemble Approach for the Prediction of Diabetes Mellitus Using a Soft Voting Classifier with an Explainable AI.

Diabetes is a chronic disease that continues to be a primary and worldwide health concern since the health of the entire population has been affected by it. Over the years, many academics have attempted to develop a reliable diabetes prediction model using machine learning (ML) algorithms. However, these research investigations have had a minimal impact on clinical practice as the current studies focus mainly on improving the performance of complicated ML models while ignoring their explainability to clinical situations. Therefore, the physicians find it difficult to understand these models and rarely trust them for clinical use. In this study, a carefully constructed, efficient, and interpretable diabetes detection method using an explainable AI has been proposed. The Pima Indian diabetes dataset was used, containing a total of 768 instances where 268 are diabetic, and 500 cases are non-diabetic with several diabetic attributes. Here, six machine learning algorithms (artificial neural network (ANN), random forest (RF), support vector machine (SVM), logistic regression (LR), AdaBoost, XGBoost) have been used along with an ensemble classifier to diagnose the diabetes disease. For each machine learning model, global and local explanations have been produced using the Shapley additive explanations (SHAP), which are represented in different types of graphs to help physicians in understanding the model predictions. The balanced accuracy of the developed weighted ensemble model was 90% with a F1 score of 89% using a five-fold cross-validation (CV). The median values were used for the imputation of the missing values and the synthetic minority oversampling technique (SMOTETomek) was used to balance the classes of the dataset. The proposed approach can improve the clinical understanding of a diabetes diagnosis and help in taking necessary action at the very early stages of the disease.

Diabetes Prediction using Machine Learning: A Review

Abstract: Prediction of diabetes using machine learning algorithms has been thoroughly studied by several researchers in the past. Finding, critically assessing, and combining the data of all pertinent, high-quality individual research is what prompted us to conduct this assessment of multiple diabetes prediction models. Analysis of several writers' expertise on diabetes prediction systems is presented in this publication. This study on diabetes prediction models aimed to discover the best strategies for selecting and synthesising the many studies of high quality. The majority of medical data is nonlinear, correlation-structured, and complicated, making it difficult to analyse. The use of machine learning-based techniques in healthcare and medical imagery has been ruled out.

Diabetes Prediction Model Comparison between XgBoost and SVM Algorithms

Diabetes is a global health epidemic. It increases the danger of cardiovascular disease by fourfold in women and around twice in men. ‘Diabetes’ is an umbrella term for a number of different subtypes of the disease. The most common are Type 1 Diabetes Mellitus (T1DM) and Type 2 Diabetes Mellitus (T2DM). Compared to men, women are also at a greater risk of retinopathy and neuropathy from diabetes. Pregnancy may worsen pre-existing conditions and lead to significant blindness. It also aggravates pre-existing kidney diseases. Elderly women with type 2 diabetes mellitus (T2DM) and end-stage renal disease have a significantly higher risk of death than men with similar diseases. Women with diabetes have higher chances of suffering a stroke in comparison to women without it. Women are also more likely to develop depression compared to men. The modeling of support vector machines may additionally be a promising classification technique for identifying women among the population with common diseases like polygenic disorder and pre-diabetes. We use different algorithms for classification, XGBoost based on SVM with GridSearchCV predict results with 83.5% accuracy.

The Association Between Plant-Based Diet Indices and Obesity and Metabolic Diseases in Chinese Adults: Longitudinal Analyses From the China Health and Nutrition Survey.

BackgroundA wide range of health benefits are associated with consuming a diet high in plant-based foods. Diet quality can be accurately assessed using plant-based diet indices, however there is inadequate evidence that plant-based diet indices are linked to obesity, hypertension, and type 2 diabetes (T2D), especially in Chinese cultures who have traditionally consumed plant-rich foods.MethodsThe data came from the China Nutrition and Health Survey. Overall, 11,580 adult participants were enrolled between 2004 and 2006 and followed up until 2009 or 2015 (follow-up rate: 73.4%). Dietary intake was assessed across three 24-h recalls, and two plant-based dietary indices [overall plant-based diet indice (PDI) and healthy plant-based diet indice (hPDI)] were calculated using China Food Composition Code and categorized into quintiles. The study's endpoints were overweight/obesity, hypertension, and T2D. The Hazard ratio (HR) and dose-response relationship were assessed using the Cox proportional risk model and restricted cubic splines. The areas under the curve of the receiver operating characteristic curve analyses were used to evaluate the predictive performance of the PDI and hPDI.ResultsDuring the median follow-up period of more than 10 years, 1,270 (33.4%), 1,509 (31.6%), and 720 (11.5%) participants developed overweight / obesity, hypertension, and T2D, respectively. The higher PDI score was linked with a reduced risk of overweight/obesity [HR: 0.71 (95% CI: 0.55–0.93), P-trend <0.001], hypertension [HR: 0.63 (95% CI: 0.51–0.79), P-trend <0.001], and T2D [HR: 0.79 (95% CI: 0.72–0.87), P-trend <0.001]. The hPDI score was inversely associated with overweight/obesity [HR: 0.79 (95% CI: 0.62–0.98), P-trend = 0.02] and T2D [HR: 0.84 (95% CI: 0.75–0.93), P-trend = 0.001]. In the aged <55-year-old group, subgroup analysis indicated a significant negative association between PDI/hPDI and overweight/obesity, hypertension, and T2D. There was no significant difference in the areas under the curve of the fully adjusted obesity, hypertension, and diabetes prediction models between PDI and hPDI.ConclusionThe PDI and hPDI scores were very similar in application in Chinese populations, and our findings highlight that adherence to overall plant-based diet index helps to reduce the risk of T2D, obesity, and hypertension in Chinese adults who habitually consume plant-based foods, especially for those aged <55 year. Further understanding of how plant-based diet quality is associated with chronic disease will be needed in the future, which will help develop dietary strategies to prevent diabetes, hypertension, and related chronic diseases.

Survey Paper on Diabetes Risk Prediction using Machine Learning Algorithm

Diabetes Mellitus (DM) is a chronic, lifelong metabolism disorder. It affects the ability of the body system to use the energy found in food. The improper management of the disease will lead to Heart disease, kidney disease, eye disease, nerve disease and pregnancy complications. Classification model helps physicians to improve their prognosis, diagnosis or treatment planning procedures. Big Data Analytics plays an significant role in healthcare industries. Healthcare industries have large volume databases. Using big data analytics one can study huge datasets and find hidden information, hidden patterns to discover knowledge from the data and predict outcomes accordingly. In existing method, the classification and prediction accuracy is not so high. In this paper, we have proposed a diabetes prediction model for better classification of diabetes which includes few external factors responsible for diabetes along with regular factors like Glucose, BMI, Age, Insulin, etc. Classification accuracy is boosted with new dataset compared to existing dataset. Further with imposed a pipeline model for diabetes prediction intended towards improving the accuracy of classification.

Dietetics Prediction System Using Machine Learning

Diabetes is a severe disease that can strike at any time and affect a large number of people. Age, obesity, sedentary lifestyle, poor diet, and high blood pressure are just few of the factors that contribute to the development of type 2 diabetes. There are a number of health problems that are more common among diabetics than in the general population. Patients with diabetes are currently being diagnosed and treated using a variety of diagnostic methods, including blood testing, urine tests, and more. In the healthcare industry, big data analytics is essential. The healthcare industry has a colossal amount of data stored in databases. Using big data analytics, users can acquire insight and make predictions about the future by examining large datasets and uncovering hidden information and trends. The current method isn't very good at classifying and forecasting. To better classify diabetes, we present a diabetes prediction model in this article that incorporates a few extrinsic parameters that cause diabetes, as well as regular components such as glucose, creatinine ratio, urea, fasting lipid profile, body mass index, age, insulin, and so on. Both datasets, each with eight variables, were subjected to the identical tests. The accuracy of a dataset with 12 variables is higher, so the conclusion is that the more information we have, the more accuracy we can attain.

Developing a simple and practical decision model to predict the risk of incident type 2 diabetes among the general population: The Di@bet.es Study

AimsTo develop a simple multivariate predictor model of incident type 2 diabetes in general population. MethodsParticipants were recruited from the Spanish Di@bet.es cohort study with 2570 subjects meeting all criteria to be included in the at-risk sample studied here. Information was collected using an interviewer-administered structured questionnaire, followed by physical and clinical examination. CHAID algorithm, which collects the information of individuals with and without type 2 diabetes, was used to develop a decision tree based type 2 diabetes prediction model. Results156 individuals were identified as having developed type 2 diabetes (6.5% incidence). Fasting plasma glucose (FPG) at the beginning of the study was the main predictive variable for incident type 2 diabetes: FPG ≤ 92 mg/dL (ref.), 92–106 mg/dL (OR = 3.76, 95%CI = 2.36–6.00), > 106 mg/dL (OR = 13.21; 8.26–21.12). More than 25% of subjects starting follow-up with FPG levels > 106 mg/dL developed type 2 diabetes. When FPG <106 mg/dL, other variables (fasting triglycerides (FTGs), BMI or age) were needed. For levels ≤ 92 mg/dL, higher FTGs levels increased risk of incident type 2 diabetes (FTGs > 180 mg/dL, OR = 14.57; 4.89–43.40) compared with the group of FTGs ≤ 97 mg/dL (FTGs = 97–180 mg/dL, OR = 3.12; 1.05–9.24). This model correctly classified 93.5% of individuals. ConclusionsThe type 2 diabetes prediction model is based on FTGs, FPG, age, gender, and BMI values. Utilizing commonly available clinical data and a simple blood test, a simple tree diagram helps identify subjects at risk of developing type 2 diabetes, even in apparently low risk subjects with normal FPG.

An Optimization-Based Diabetes Prediction Model Using CNN and Bi-Directional LSTM in Real-Time Environment

Diabetes is a long-term illness caused by the inefficient use of insulin generated by the pancreas. If diabetes is detected at an early stage, patients can live their lives healthier. Unlike previously used analytical approaches, deep learning does not need feature extraction. In order to support this viewpoint, we developed a real-time monitoring hybrid deep learning-based model to detect and predict Type 2 diabetes mellitus using the publicly available PIMA Indian diabetes database. This study contributes in four ways. First, we perform a comparative study of different deep learning models. Based on experimental findings, we next suggested merging two models, CNN-Bi-LSTM, to detect (and predict) Type 2 diabetes. These findings demonstrate that CNN-Bi-LSTM surpasses the other deep learning methods in terms of accuracy (98%), sensitivity (97%), and specificity (98%), and it is 1.1% better compared to other existing state-of-the-art algorithms. Hence, our proposed model helps clinicians obtain complete information about their patients using real-time monitoring and can check real-time statistics about their vitals.

PREDICTION OF GESTATIONAL DIABETES AND HYPERTENSION BASED ON PREGNANCY EXAMINATION DATA

Gestational diabetes mellitus and hypertension are two common pregnancy complications, which seriously threaten the life safety of pregnant women and adversely affect the growth and development of the fetus. Therefore, it is of great significance to detect and prevent hypertension and diabetes at an early stage of pregnancy. Each pregnant woman will undergo multiple tests at different gestational weeks. This progress produces lots of pregnancy examination data. These data can reflect the dynamic changes of pregnant women’s health indicators during pregnancy. This study aims to establish gestational diabetes and hypertension prediction model with a machine learning method based on real pregnancy examination data from the hospital. We use Logistic Regression, XGBoost, LightGBM, and Neural Network Model based on LSTM to do the prediction, respectively, and compare the performance. We check the prediction accuracy at different stages of pregnancy. We found that with pregnancy examination data at all gestational weeks, the predictive AUCs for diabetes and hypertension can reach 0.92 and 0.87, respectively. At 16th gestational week, the AUCs are 0.68 for diabetes and 0.70 for hypertension. We extract the checking items which are most important and get a simplified model with a modest reduction in predictive accuracy. This study demonstrates that based on several routine pregnancy examination items we can establish a machine learning model to detect and predict gestational diabetes and hypertension. This can be used as a diagnostic aid and is conducive to early prevention and treatment.

Diabetes mellitus prediction and diagnosis from a data preprocessing and machine learning perspective

Background and ObjectiveDiabetes mellitus is a metabolic disorder characterized by hyperglycemia, which results from the inadequacy of the body to secrete and respond to insulin. If not properly managed or diagnosed on time, diabetes can pose a risk to vital body organs such as the eyes, kidneys, nerves, heart, and blood vessels and so can be life-threatening. The many years of research in computational diagnosis of diabetes have pointed to machine learning to as a viable solution for the prediction of diabetes. However, the accuracy rate to date suggests that there is still much room for improvement. In this paper, we are proposing a machine learning framework for diabetes prediction and diagnosis using the PIMA Indian dataset and the laboratory of the Medical City Hospital (LMCH) diabetes dataset. We hypothesize that adopting feature selection and missing value imputation methods can scale up the performance of classification models in diabetes prediction and diagnosis. MethodsIn this paper, a robust framework for building a diabetes prediction model to aid in the clinical diagnosis of diabetes is proposed. The framework includes the adoption of Spearman correlation and polynomial regression for feature selection and missing value imputation, respectively, from a perspective that strengthens their performances. Further, different supervised machine learning models, the random forest (RF) model, support vector machine (SVM) model, and our designed twice-growth deep neural network (2GDNN) model are proposed for classification. The models are optimized by tuning the hyperparameters of the models using grid search and repeated stratified k-fold cross-validation and evaluated for their ability to scale to the prediction problem. ResultsThrough experiments on the PIMA Indian and LMCH diabetes datasets, precision, sensitivity, F1-score, train-accuracy, and test-accuracy scores of 97.34%, 97.24%, 97.26%, 99.01%, 97.25 and 97.28%, 97.33%, 97.27%, 99.57%, 97.33, are achieved with the proposed 2GDNN model, respectively. ConclusionThe data preprocessing approaches and the classifiers with hyperparameter optimization proposed within the machine learning framework yield a robust machine learning model that outperforms state-of-the-art results in diabetes mellitus prediction and diagnosis. The source code for the models of the proposed machine learning framework has been made publicly available.