Feasible Prediction of Diabetes in Pregnant Woman and Neonatal Mellitus in New Born Child using Machine Learning

. Diabetes during pregnancy is a major source of health problems in unborn infants and their moms. Because gestational diabetes can develop to permanent diabetes, ML is an important method for predicting the likelihood of such progression based on the given features. Although the current study may predict lifelong diabetes in pregnant women, it cannot predict the likelihood of neonatal diabetes. As a result, new characteristics are required to improve the forecasting of neonatal mellitus and provide the most accurate and feasible diabetes persistence results in pregnant women. Python scripting and the application of Machine Learning methods such as SVM, KNN, and LR can assist in achieving this aim. The preprocessing ML dataset focusing on Diabetes from the Pima Indian diabetes database collected through Kaggle. In addition, two new attributes were added to the paper's dataset. According to research, machine learning models using characteristics like SVM and decision trees may successfully predict the risk of diabetes in pregnant women. Various factors have been used to predict the beginning of this condition during pregnancy.


Introduction
Machine learning algorithms utilize previous data to discover patterns and make predictions about new data.In the case of diabetes prediction, machine learning algorithms may use data from electronic health records, clinical notes, test findings, and patient demographics to identify people at risk of acquiring diabetes.Several researches have looked at the use of ML algorithms to predict diabetes in pregnant women and neonatal diabetes in babies.One research, for example, employed ML algorithms to foresee the risk of gestational diabetes in pregnant women based on demographic and clinical variables.In terms of predicting gestational diabetes, the algorithm has an accuracy of 89.4%.
These algorithms make use of a many of data sources, such as clinical factors, Laboratory tests, and Demographic data.Using these data sets, the algorithms can accurately identify women who are at high risk of having GDM.NDM (neonatal diabetes mellitus) is an uncommon type of diabetes that develops within the first six months of life.Genetic mutations that impair insulin synthesis or function create the disease.NDM is a diverse disorder with different genetic alterations in different people.NDM can be temporary, resolving within a few months of birth, or persistent, necessitating lifetime care.
Gathering relevant data from pregnant women and new-borns, including medical history, glucose levels, BMI, and other factors.This data is preprocessed, selecting important features and ensuring data quality.ML algorithms such as LR, DT, or NN are then trained using this data to establish patterns and relationships between the features and diabetes outcomes.The trained model is evaluated using validation data, and once deemed satisfactory, it can be utilized to predict diabetes risk in pregnant women and neonatal diabetes in new-borns which is very feasible.These predictions aid in identifying high-risk cases and enabling timely interventions and monitoring by healthcare professionals.

Challenges and pitfalls
While the application of ML algorithms for foreseeing diabetes in pregnant women and neonatal diabetes mellitus shows considerable potential, there are various issues that must be solved in order for these models to be accurate and useful.One of the most difficult aspects of forecasting diabetes in pregnant women is the condition's heterogeneity.Gestational diabetes mellitus (GDM) is a complicated illness impacted by a variety of variables such as maternal age, BMI, ethnicity, and diabetes family history.Furthermore, GDM is a changing disease that can occur throughout pregnancy.Because of this complexity, developing reliable prediction models that account for all of these elements and changes is difficult.

Existing methods
One frequent strategy is to train machine learning algorithms using clinical data such as glucose levels, blood pressure, and body mass index, as well as demographic and medical history information.These algorithms, which include decision trees, support vector machines, and neural networks, may find patterns and correlations in data to produce diabetes risk predictions.Furthermore, researchers have investigated the use of genetic data and biomarkers to improve prediction accuracy.These techniques are intended to promote early identification and intervention, resulting in improved health outcomes for both pregnant women and new born infants.The paper [8] discussed the role of Intelligent Decision Support Systems (IDSS) in Healthcare Monitoring, especially for heart disease.Results claimed that IDSS enhances decision-making functionalities in uncertain healthcare scenarios, there by significantly improving the monitoring and remedial activities.The paper [9] explores the distinct ML applications in predicting heart attacks using patient health records.It compares Random Forest and CNN methods, and findings showed that Random Forest's better performance in terms of accuracy.Authors [10] highlighted the significance of ML in prediction, pattern recognition and error reduction across diverse fields, emphasizing the impact of AI in broad domain.Authors [11] suggested data mining techniques to predict disease-prevalence based on symptoms in healthcare data.The appropriate prediction helps healthcare organizations avoid drug shortages and further ensures timely treatment of patients.Authors [12] suggested the different issues and challenges related to IoT based e-healthcare.The results showed that prediction with random forest could reach the highest accuracy (ACC = 0.8084) when all the attributes were used.[3] SVM Support vector machine classifier demonstrated the highest accuracy, 78.1% predicting the development of gestational diabetes based.
The most accurate method was determined to be logistic regression and this resulted in an accuracy of 78.01 percent.
[5] Random Forest The overall predictive accuracy of the F1 model was 93.10%, the predictive accuracy of GDM-positive cases 37.10%.The corresponding values for the F2 model 88.70%, and 79.44%. [6] The ADA prognostic model, consisting of age and BMI at booking Selective screening of women for GDM using the ADA model with a risk threshold of 3% gave 93% sensitivity for identification of women with GDM with a 27% reduction in the number of OGTTs required. [7] Stepwise Logistic Regression (age, previous GDM, family history of type 2 diabetes, waist: height and neck: thigh ratios) provided an area under the curve of 0.71 (95%CI 0.68-0.74) 3 Problem statement and objectives

Problem statement
Diabetes is a chronic condition that affects millions of individuals all over the world and can have major health repercussions, particularly during pregnancy.Gestational Diabetes Mellitus (GDM) is a frequent condition that can result in neonatal mellitus in neonates.Early identification and treatment of GDM are critical for avoiding negative effects for both the mother and the child.Current GDM detection approaches rely heavily on blood glucose tests and clinical risk factors.However, these strategies may not be precise enough to forecast GDM in all circumstances.Machine learning (ML) has showed potential in predicting GDM by analysing massive datasets and finding risk variables that traditional approaches may miss.The problem statement for utilising machine learning to predict diabetes in pregnant women and neonatal mellitus in newborns entails establishing accurate and reliable ML models that gives a feasible prediction of the GDM in pregnant women and neonatal mellitus.

Objective
The paper's goal is to use machine learning techniques to forecast the risk of permanent diabetes in pregnant women and Neonatal Mellitus mellitus in newborn children.By developing machine learning models and providing a user interface for on-spot data calculation, the paper seeks to facilitate early detection and prevention of diabetes in these populations.Large datasets will be analysed using machine learning algorithms to uncover patterns and risk variables related with persistent diabetes in pregnant women and Neonatal Mellitus diabetes in newborns.By training these models on comprehensive and diverse datasets, the goal is to ensure their reliability and generalizability in real-world scenarios.The paper aims to develop a feasible predictive model using machine learning techniques to identify the risk of diabetes in pregnant women and neonatal mellitus in new born children.The paper involves collecting comprehensive data, including medical history, glucose levels, BMI, and other relevant factors from pregnant women and new borns.This data is then preprocessed and used to train machine learning algorithms such as logistic regression, decision trees, or neural networks.The trained model will enable early detection of diabetes risk, allowing for timely interventions and personalized care plans.The paper also focuses on risk stratification, cost-effectiveness, and generating valuable insights for further research and understanding of diabetes in these populations.

Architecture of the proposed work
1.The Raw Dataset serves as the starting point for the architecture.This dataset is specifically curated to contain a collection of data of various females with symptoms of diabetes.The data is typically sourced from various real-world scenarios or collected under specific conditions.2. Data Preprocessing is an essential step in preparing the blurred images from the RealBlur dataset for training and evaluation.In this step, several operations are performed to transform the raw image data into a suitable format for feeding into a , 010 (2023) E3S Web of Conferences ICMPC 2023 https://doi.org/10.1051/e3sconf/20234300104747 430 neural network.The main operations typically include conversion to tensors and resizing 3.After performing various data preprocessing techniques on the raw data set we get a clean dataset that is further used for the paper.4. Model Training is a Crucial step in the paper as the various models like Naive Bayes, K-Nearest Neighbours, Random Forest, Logistic Regression, SVM are used to train the dataset and the best of the mentioned is used for final execution.5. Predictions are drawn from the training process to know that which model is better in the prediction.6.In this step, the model is evaluated using the preprocessed testing data, which consist of diabetes data of females and hence the best model is drawn out of the five.7.Here the user can input the data he has and then the algorithm runs in the background to calculate whether the user has diabetes or not.8.Here the algorithm generates output and displays prompts either "You are diabetic" or "You are not diabetic".Along with the accuracy of the calculation.(mmol/L) are the most common units of measurement.Elevated glucose levels might be a sign of poor glucose tolerance or diabetes.
Blood Pressure-Typically, this characteristic comprises two values: systolic and diastolic blood pressure.When the heart contracts, systolic blood pressure indicates the pressure in the arteries, whereas diastolic blood pressure shows the pressure when the heart is at rest.Hypertension (high blood pressure) is a risk factor for gestational diabetes.
BMI (Body Mass Index)-BMI is a measure of body fat that is dependent on a person's height and weight.It is computed by dividing the weight in kilogrammes by the height in metres squared.BMI indicates if a patient is underweight, normal weight, overweight, or obese.Obesity is a major risk factor for developing gestational diabetes.The proposed method of predicting diabetes in pregnant women using machine learning holds significant potential in advancing healthcare outcomes in this domain which is feasible.By leveraging machine learning algorithms and predictive modelling techniques, this approach aims to provide a feasible prediction of diabets and risk assessment, enabling timely interventions and personalized care.One of the key advantages of this method is its potential to improve maternal and neonatal health outcomes by identifying high-risk cases accurately.Machine learning models can analyze large amounts of data from pregnant women, including medical history, lifestyle factors, and various clinical parameters, to identify patterns and generate predictive insights.

Conclusion and future enhancements
Feasible prediction of diabetes in pregnant women and neonatal mellitus in new born children using machine learning revolves around developing predictive models to predict the risk of gestational diabetes in women expectant mothers and the likelihood of neonatal diabetes in new born.By leveraging machine learning algorithms and analysing various factors such as medical history, clinical parameters, lifestyle, and genetic predisposition, these models aim to provide early detection and personalized risk assessment.The goal is to enable healthcare professionals to intervene proactively, implement preventive measures.The potential of machine learning to predict gestational diabetes and neonatal mellitus is promising and encompasses several potential advancements and areas of exploration.Here are some key aspects that offer significant opportunities for further development in terms of accuracy and longitudinal analysis.

1 . 3 .
Early detection allows timely interventions and improves outcomes.2. Personalized care plans based on individual risk factors enhance treatment effectiveness.Risk stratification optimizes healthcare delivery and resource allocation.4. Cost savings through reduced complications and hospitalizations. 5. Insights gained contribute to further research and understanding of diabetes.

Table 1 .
Summary of approaches.