OPTIMIZING EARLY DIABETES DETECTION: A MACHINE LEARNING APPROACH

Abstract

The present invention relates to a diabetes prediction method and system based on the field of machine learning. The system for early detection of diabetes mellitus, comprising: receiving and interpreting dataset representing of a patient having been diagnosed with diabetes mellitus to identify patterns and trends; pre-processing and identifying inconsistent data in this stage of the model to produce more precise and accurate findings; selecting and model building from among a plurality of models in a machine learning system that has been trained for detecting diabetes; and analyzing performance of the classification models test dataset using a confusion matrix; the machine learning model is trained using training data that comprises dataset representing of a plurality of training patients; each model in the machine learning system is trained for diabetes prediction is built using number of machine learning algorithms for predicting diabetes; determining a prediction of diabetes mellitus events for the patient by processing the medical records of the patient using the machine learning model comparing the predicted rate to a predetermined value. The system for early detection of diabetes mellitus, provides a powerful method for the early identification of diabetic disease along with machine learning and AI-based interface that assists to monitor and manage the complication of diabetes mellitus at an earlier stage.

Specification

Description:Technical Field of the Invention
The invention belongs to an innovative method for Diabetes Mellitus prediction and diagnosis based on automated predictive model. In particular, the present invention relates to a novel method based on a machine learning built model, integrating diabetes-related datasets. The method is capable of detecting and monitoring for early Diabetes Mellitus diagnosis, disease prediction and prognosis.
Background of the Invention
Diabetes is a chronic metabolic disorder characterized by high blood glucose levels, which over time can cause significant damage to vital organs such as the heart, blood vessels, eyes, kidneys, and nerves. The most common form, type 2 diabetes, typically affects adults and occurs when the body either fails to produce enough insulin or becomes resistant to it. Over the past three decades, the incidence of type 2 diabetes has sharply risen in countries across all income levels. Type 1 diabetes, also known as juvenile or insulin-dependent diabetes, is a chronic condition where the pancreas produces little to no insulin. Access to affordable medications, such as insulin, is crucial for those with diabetes to manage the condition and survive. One of the challenges in managing diabetes is making accurate predictions based on clinical data. In the field of information technology, machine learning is an emerging discipline that focuses on how machines can learn from experience. Machine learning is a subclass of artificial intelligence in the field of computer science that often uses statistical methods and techniques to train computers the ability to "learn" and identify the patterns and trends. Training the computers in machine learning progressively improve performance on a specific task with data without being explicitly programmed [Durga, P. and Sudhakar, T., 2022. An Analysis of various Machine Learning Techniques for Predicting Diabetes in its Early Stages. Journal of Pharmaceutical Negative Results, pp.2030-2038; Althobaiti, T., Althobaiti, S. and Selim, M.M., 2024. An optimized diabetes mellitus detection model for improved prediction of accuracy and clinical decision-making. Alexandria Engineering Journal, 94, pp.311-324; Wee, B.F., Sivakumar, S., Lim, K.H., Wong, W.K. and Juwono, F.H., 2024. Diabetes detection based on machine learning and deep learning approaches. Multimedia Tools and Applications, 83(8), pp.24153-24185]. The present invention aims to develop a system that can accurately detect diabetes early by integrating multiple machine-learning techniques. The project also proposes a powerful approach to the early diagnosis of diabetes.
Objects of the Invention
The main object of the present invention is to provide an innovative method for Diabetes Mellitus prediction and diagnosis based on automated computer implemented predictive model.
Yet another object of the present invention is to provide a system to manage and monitor the early diagnosis of diabetes.
Another object of the present invention is to provide method for the prediction of pre-diabetes or diabetes disorder.
Still another object of the present invention is to provide an expeditious and cost-effective AI-based system for prediction of diabetes for advantageous in healthcare management system. Summary of the Invention
The present invention provides a system for early detection of diabetes mellitus, comprising: a) receiving and interpreting dataset representing of a patient having been diagnosed with diabetes mellitus to identify patterns and trends; b) pre-processing and identifying inconsistent data in this stage of the model to produce more precise and accurate findings; c) selecting and model building from among a plurality of models in a machine learning system that has been trained for detecting diabetes mellitus; and d) evaluating and analyzing performance of the classification models test dataset using a confusion matrix; wherein the machine learning model is trained using training data that comprises dataset representing of a plurality of training patients and corresponding rates of diabetes mellitus events for the respective training patients; wherein each model in the machine learning system is trained for diabetes prediction is built using number of machine learning algorithms for predicting diabetes; wherein determining a prediction of diabetes mellitus events for the patient by processing the medical records of the patient using the machine learning model comparing the predicted rate to a predetermined value in response to determining diabetes mellitus. In an embodiment, the present invention provides a system for early detection of diabetes mellitus, wherein the experiments are conducted using a python programming language with the help of powerful libraries functions; wherein the dataset consists of 8 attributes and 768 observations. In an embodiment, the present invention provides a system for early detection of diabetes mellitus, wherein the parameters applied are sensitivity or recall, accuracy, precision, F1-score, and duration (milliseconds), the confusion matrix to the observed results. In an embodiment, the present invention provides a system for early detection of diabetes mellitus, wherein the use of various machine learning algorithms on the dataset shows the accuracy results with the maximum accuracy of 96% provided by logistic regression. In an embodiment, the present invention provides a system for early detection of diabetes mellitus, wherein the comparison of Performance Metrics for different ML Algorithms shows LR Recall 96, Accuracy 96, Precision 95, F1-Score 96; RF Recall 78, Accuracy 91, 78, 79; DT Recall 81, Accuracy 86, Precision 79, F1-Score 77 and KNN Recall 89, Accuracy 90, Precision 99, F1-Score 97. In an embodiment, the present invention provides a system for early detection of diabetes mellitus, wherein the accuracies of different ML models on Diabetes Dataset using Algorithms show Logistic Regression 96%, Random Forest 91%, Decision Tree 86% and KNN 90%. In an embodiment, the present invention provides a system for early detection of diabetes mellitus, wherein the present invention provides a powerful method for the early identification of diabetic disease. In an embodiment, the present invention provides a system for early detection of diabetes mellitus, wherein the present invention provides machine learning and AI-based knowledge interface that assists patients to monitor and manage the complication of diabetes mellitus at an earlier stage.
Brief Description of drawings
In the drawings accompanying the specification, Figure 1 shows role of Machine Learning in Disease Prediction.
In the drawings accompanying the specification, Figure 2 shows Proposed Work and an architecture schematic for the diabetes prediction model.
In the drawings accompanying the specification, Figure 3 shows Comparison of Performance Metrics for Different Machine Learning Algorithms.
In the drawings accompanying the specification, Figure 4 shows Comparison of various machine learning algorithms based on accuracies.
In the drawings accompanying the specification, Figure 5 shows Graph Representation of Diabetes Prediction.
Detailed description of the Invention
The present invention provides a system and method for early detection of diabetes mellitus, comprising: a) receiving and interpreting dataset representing of a patient having been diagnosed with diabetes mellitus to identify patterns and trends; b) pre-processing and identifying inconsistent data in this stage of the model to produce more precise and accurate findings; c) selecting and model building from among a plurality of models in a machine learning system that has been trained for detecting diabetes mellitus; and d) evaluating and analyzing performance of the classification models test dataset using a confusion matrix; wherein the machine learning model is trained using training data that comprises dataset representing of a plurality of training patients and corresponding rates of diabetes mellitus events for the respective training patients; wherein each model in the machine learning system is trained for diabetes prediction is built using number of machine learning algorithms for predicting diabetes; wherein determining a prediction of diabetes mellitus events for the patient by processing the medical records of the patient using the machine learning model comparing the predicted rate to a predetermined value in response to determining diabetes mellitus. The system for early detection of diabetes mellitus, wherein the experiments are conducted using a python programming language with the help of powerful libraries functions; wherein the dataset consists of 8 attributes and 768 observations. The system for early detection of diabetes mellitus, wherein the parameters applied are sensitivity or recall, accuracy, precision, F1-score, and duration (milliseconds), the confusion matrix to the observed results. The system for early detection of diabetes mellitus, wherein the use of various machine learning algorithms on the dataset shows the accuracy results with the maximum accuracy of 96% provided by logistic regression. The system for early detection of diabetes mellitus, wherein the comparison of Performance Metrics for different Machine Learning Algorithms shows LR Recall 96, Accuracy 96, Precision 95, F1-Score 96; RF Recall 78, Accuracy 91, 78, 79; DT Recall 81, Accuracy 86, Precision 79, F1-Score 77 and KNN Recall 89, Accuracy 90, Precision 99, F1-Score 97. The system for early detection of diabetes mellitus, wherein the accuracies of different ML models on Diabetes Dataset using Algorithms show Logistic Regression 96%, Random Forest 91%, Decision Tree 86% and KNN 90%. The system for early detection of diabetes mellitus, wherein the present invention provides a powerful method for the early identification of diabetic mellitus. The system for early detection of diabetes mellitus, wherein the present invention provides machine learning and AI-based knowledge interface that assists patients to monitor and manage the complication of diabetes mellitus at an earlier stage of disease.
METHODOLOGY
Disease Prediction Using Machine Learning
Machine learning techniques are frequently employed to forecast diabetes, and they produce better outcomes. One of the common machine learning techniques in the medical industry that has excellent classification capability is the decision tree. Many decision trees are produced from the random forest. A machine learning technique that has gained popularity recently and offers superior performance in several areas is the neural network.
Fig 1 shows Role of ML in Disease Prediction.
Models for Supervised Learning and Prediction
Algorithms are used in the construction of predictive models and are supervised during learning. A prediction model projects the values that are missing using other dataset values. Using a set of input data and a set of output data, a supervised learning technique builds a model that can make precise predictions about how a new dataset would react. DT, ANN, Instance-based Learning, Bayesian Method, and Ensemble Method are a few examples of supervised learning techniques. These machine learning techniques are highly sought after.
Descriptive models and unsupervised learning
Unsupervised learning is used to create descriptive models. In this model, the inputs are known, but the output is not. Transactional data is where unsupervised learning is most commonly applied. The clustering techniques used in this method include k-Means clustering and k-Medians clustering.
Semi-supervised Learning
On the training dataset, the semi-supervised learning technique uses both labeled and unlabeled data. Techniques for classification and regression fall under semi-supervised learning. Regression methods include linear regression and logistic regression, for instance.
Dataset
PIMA and Indian diabetes datasets were downloaded for this work via Kaggle. We have 768 observations and 8 variables in the PIMA Indian Diabetes dataset, which includes an outcome of 0 (or) 1. A score of 0 indicates that a person does not have diabetes, whereas a score of 1 indicates that they do. The diabetes dataset for Pima Indians is split in two parts: 80% and 20%. 80 percent of the data are from training, and 20 percent are from testing. In this study, we use the Python programming language to create classification techniques like the Random Forest tree and the Logistic Regression algorithm.
Proposed System
In the suggested method, supervised learning from machine learning is employed to produce precise results for chronic conditions like diabetes. There are two sections to the dataset. 1) Training part 2) Testing portion. The model is created by applying the algorithm to the training set, which is then used to produce the model for the test, after which its performance is calculated. In the outlined study, we used Python programming to apply the Logistic Regression and Random Forest techniques to the Pima Indian diabetes dataset.
Figure 2 shows Proposed Work and an architecture schematic for the diabetes prediction model. There are five distinct modules in this model. These modules consist of :
Dataset Collection
Data Pre-processing
Model Building
Performance Analysis
Dataset Collection
Data collection and interpretation are covered in this module in order to identify patterns and trends that may be used to forecast outcomes and assess them. So here is a description of the data set. There are 8 attributes and 768 observations in this diabetes dataset.
Data Pre-processing
Inconsistent data are handled in this stage of the model to produce more precise and accurate findings. Missing values can be found in this dataset. Due to the fact that certain attributes, such as age, skin thickness, blood pressure, and BMI, cannot have zero values, we imputed missing values for these attributes. The dataset is then scaled to normalize each value.
Model Building
This is the most crucial stage, during which a model for diabetes prediction is built. In this, we've created a number of machine learning algorithms for predicting diabetes. These algorithms include K-Nearest Neighbor, Logistic Regression, Decision Tree, and Random Forest Classifier.
Performance Analysis
The performance analysis is mainly focused on calculated by using a confusion matrix
Confusion Matrix
The performance of the classification models for a certain set of test data is evaluated using a matrix called the confusion matrix. Only after the true values of the test data are known can it be determined.
True Negative: The model predicted No, and the actual or true value likewise indicated No.
True Positive: The model correctly predicted yes, and the outcome matched that prediction.
False Negative: This error is sometimes referred to as a Type-II error when the model predicted no but the actual value was yes.
False Positive: Although the model predicted "Yes," the actual result was "No." Another name for it is a Type-I error.
Classification Accuracy: This is a crucial factor in figuring out how accurate a problem's classification is. It specifies how frequently the model predicts the right result. The number of accurate predictions made by the classifier divided by the total number of predictions made by the classifiers can be used to compute it. The following is the formula:
Accuracy=(TP+TN)/(TP+TN+FP+FN)
Precision: It can be characterized as the number of accurate outputs produced by the model or as the proportion of accurately anticipated positive classifications that actually occurred. Using the formula below, it can be calculated:
Precision=TP/(TP+FP)
Recall: It is considered as the positive classes that are not included in the total, as our model accurately anticipated. There must be a significant recall.
Recall=TP/(TP+ FN)
F-measure: It is challenging to compare two models if one has a high recall and a poor precision. F-score can therefore be used for this purpose. This score enables us to simultaneously assess recall and precision. If the recall and precision are equal, the F-score is at its highest. Using the formula below, it can be calculated:
F1-Score=2* ((Precision*Recall))/((Precision+Recall))
 
Logistic Regression
The dependent variable in a logistic regression model is categorical, specifically binary, meaning that it can only take the values "0" and "1," which reflect outcomes like pass/fail, win/lose, alive/dead, or healthy/sick. The majority of medical areas, social sciences, and machine learning all use logistic regression. For instance, logistic regression was initially used to generate the Trauma and Injury Severity Score (TRISS), which is frequently used to predict death in injured patients. Logistic regression has been used to generate numerous more medical measures that are used to evaluate a patient's severity. The method is also applicable to engineering, particularly when determining the likelihood that a certain process, system, or product may fail. The ability to estimate a customer's likelihood to buy a product or cancel a subscription is frequently employed in marketing applications. A business application is about estimating the likelihood of a homeowner defaulting on a mortgage. It can be used to predict a person's likelihood of choosing to be in the labor force in economics. Natural language processing employs conditional random fields, a logistic regression extension to sequential data. In this study, the presence or absence of diabetes was predicted using logistic regression using seven patient features.
Experimental Results
Experiments are conducted by using a python programming language with the help of powerful libraries functions. The dataset consists of 8 attributes and 768 observations. By using the following parameters such as sensitivity or recall, accuracy, precision, F1-score, and duration (milliseconds). Apply the confusion matrix to the observed results. Despite the use of various machine learning algorithms on the dataset, the accuracy results are as follows. The maximum accuracy, 96 percent, is provided by logistic regression.
Fig 3: Comparison of Performance Metrics for Different ML Algorithms
 
Table 1: compares the diabetes datasets used in this study and the PIMA diabetes dataset
Algorithms Recall Accuracy Precision F1-Score
LR 96 96 95 96
RF 78 91 78 79
DT 81 86 79 77
KNN 89 90 99 97
Table 2: Accuracies of Different ML models on Two datasets
Algorithms Accuracy with PIMA Data set Accuracy with Diabetes Dataset used in this paper
Logistic Regression 76% 96%
Random Forest 72% 91%
Decision Tree 74% 86%
KNN 72% 90%
Accuracy, F1-Score, Precision, and Recall are the many performance metrics being compared. Visualizing these accuracy helps us comprehend the differences between them.
Figure 4 shows Comparison of various machine learning algorithms based on accuracies.
Figure 5 shows Graph Representation of Diabetes Prediction.
Conclusion
With the advancement of computational techniques and the availability of extensive epidemiological and genetic datasets on diabetes risk, machine learning has the potential to revolutionize the prediction of diabetes risk. Early detection of diabetes is crucial for effective treatment. In this study, various machine learning algorithms were applied to a dataset, with classification achieved through different techniques, and Logistic Regression achieved the highest accuracy of 96%. This study presents a machine learning approach for predicting diabetes risk levels. The method can also support researchers in developing a precise and practical tool for healthcare providers, aiding in better decision-making regarding the patient's condition. Furthermore, this research can be expanded to predict the likelihood of non-diabetic individuals developing diabetes in the future.
References
1. Z. Punthakee, R. Goldenberg, and P. Katz, "Definition, Classification, and Diagnosis of Diabetes, Prediabetes and Metabolic Syndrome," Can. J. Diabetes, vol. 42, pp. S10-S15, 2018.
2. M. N. Piero, "Diabetes mellitus - a devastating metabolic disorder," Asian J. Biomed. Pharm. Sci., vol. 4, no. 40, pp. 1-7, 2015
3. J. Xie and Q. Wang, "Benchmarking Machine Learning Algorithms on Blood Glucose Prediction for Type I Diabetes in Comparison With Classical Time-Series Models," in IEEE Transactions on Biomedical Engineering, vol. 67, no. 11, pp. 3101-3124, Nov. 2020, doi: 10.1109/TBME.2020.2975959.
4. B. J. Lee and J. Y. Kim, "Identification of Type 2 Diabetes Risk Factors Using Phenotypes Consisting of Anthropometry and Triglycerides based on Machine Learning," in IEEE Journal of Biomedical and Health Informatics, vol. 20, no. 1, pp. 39-46, Jan. 2016, doi: 10.1109/JBHI.2015.2396520.
5. T. M. Le, T. M. Vo, T. N. Pham and S. V. T. Dao, "A Novel Wrapper-Based Feature Selection for Early Diabetes Prediction Enhanced With a Metaheuristic," in IEEE Access, vol. 9, pp. 7869-7884, 2021, doi: 10.1109/ACCESS.2020.3047942.
6. M. S. Islam, M. K. Qaraqe, S. B. Belhaouari and M. A. Abdul-Ghani, "Advanced Techniques for Predicting the Future Progression of Type 2 Diabetes," in IEEE Access, vol. 8, pp. 120537-120547, 2020, doi: 10.1109/ACCESS.2020.3005540.
7. S. Perveen, M. Shahbaz, T. Saba, K. Keshavjee, A. Rehman and A. Guergachi, "Handling Irregularly Sampled Longitudinal Data and Prognostic Modeling of Diabetes Using Machine Learning Technique," in IEEE Access, vol. 8, pp. 21875-21885, 2020, doi: 10.1109/ACCESS.2020.2968608.
8. N. Fazakis, O. Kocsis, E. Dritsas, S. Alexiou, N. Fakotakis and K. Moustakas, "Machine Learning Tools for Long-Term Type 2 Diabetes Risk Prediction," in IEEE Access, vol. 9, pp. 103737-103757, 2021, doi: 10.1109/ACCESS.2021.3098691.
9. M. Shokrekhodaei, D. P. Cistola, R. C. Roberts and S. Quinones, "Non-Invasive Glucose Monitoring Using Optical Sensor and Machine Learning Techniques for Diabetes Applications," in IEEE Access, vol. 9, pp. 73029-73045, 2021, doi: 10.1109/ACCESS.2021.3079182.
10. U. Ahmed et al., "Prediction of Diabetes Empowered With Fused Machine Learning," in IEEE Access, vol. 10, pp. 8529-8538, 2022, doi: 10.1109/ACCESS.2022.3142097.
11. N. E. Costea, E. V. Moisi and D. E. Popescu, "Comparison of Machine Learning Algorithms for Prediction of Diabetes," 2021 16th International Conference on Engineering of Modern Electric Systems (EMES), 2021, pp. 1-4, doi: 10.1109/EMES52337.2021.9484116.
12. A. Yahyaoui, A. Jamil, J. Rasheed and M. Yesiltepe, "A Decision Support System for Diabetes Prediction Using Machine Learning and Deep Learning Techniques," 2019 1st International Informatics and Software Engineering Conference (UBMYK), 2019, pp. 1-4, doi: 10.1109/UBMYK48245.2019.8965556.
13. B. Nithya and Dr. V. Ilango," Predictive Analytics in Health Care Using Machine Learning Tools and Techniques", International Conference on Intelligent Computing and Control Systems, 978-1-5386-2745-7,2017.
, Claims:We claim:
1. A system for early detection of diabetes mellitus, comprising:
a) receiving and interpreting dataset representing of a patient having been diagnosed with diabetes mellitus to identify patterns and trends;
b) pre-processing and identifying inconsistent data in this stage of the model to produce more precise and accurate findings;
c) selecting and model building from among a plurality of models in a machine learning system that has been trained for detecting diabetes mellitus; and
d) evaluating and analyzing performance of the classification models test dataset using a confusion matrix;
wherein the machine learning model is trained using training data that comprises dataset representing of a plurality of training patients and corresponding rates of diabetes mellitus events for the respective training patients;
wherein each model in the machine learning system is trained for diabetes prediction is built using number of machine learning algorithms for predicting diabetes;
wherein determining a prediction of diabetes mellitus events for the patient by processing the medical records of the patient using the machine learning model comparing the predicted rate to a predetermined value in response to determining diabetes mellitus.
2. The system for early detection of diabetes mellitus as claimed in claim 1, wherein the experiments are conducted using a python programming language with the help of powerful libraries functions; wherein the dataset consists of 8 attributes and 768 observations.
3. The system for early detection of diabetes mellitus as claimed in claim 1, wherein the applied parameters are sensitivity or recall, accuracy, precision, F1-score, and duration (milliseconds), the confusion matrix to the observed results.
4. The system for early detection of diabetes mellitus as claimed in claim 1, wherein the use of various machine learning algorithms on the dataset shows the accuracy results with the maximum accuracy of 96% provided by logistic regression.
5. The system for early detection of diabetes mellitus as claimed in claim 1, wherein the comparison of Performance Metrics for different ML Algorithms shows LR Recall 96, Accuracy 96, Precision 95, F1-Score 96; RF Recall 78, Accuracy 91, 78, 79; DT Recall 81, Accuracy 86, Precision 79, F1-Score 77 and KNN Recall 89, Accuracy 90, Precision 99, F1-Score 97.
6. The system for early detection of diabetes mellitus as claimed in claim 1, wherein the accuracies of different ML models on Diabetes Dataset using Algorithms show Logistic Regression 96%, Random Forest 91%, Decision Tree 86% and KNN 90%.
7. The system for early detection of diabetes mellitus as claimed in claim 1, wherein the present invention provides a powerful method for the early identification of diabetic disease.
8. The system for early detection of diabetes mellitus as claimed in claim 1, wherein the present invention provides machine learning and AI-based knowledge interface that assists patients to monitor and manage the complication of diabetes mellitus at an earlier stage.
Dated this 16th day of November, 2024

Documents