Diabetes Detection Using Machine Learning Project

1. Comparison of Performance of Machine Learning Algorithms for Diabetes Detection

Keywords

Diabetes detection, KNN, Machine learning, Random Forest, Support Vector Machine, XGBoost

            Forecasting of type 2 diabetes through the utilization of various ML approaches is the main objective of our study. We evaluated several methods such as Support Vector Machine, K-Nearest Neighbours, XGBoost, and Random Forest Regression. We compared these methods in terms of various metrics to find out the efficient one. Results show that, Random Forest achieved highest outcomes than others.

2. Effective Feature Selection and Soft Voting Classifier based Diabetes Detection Using Machine Learning Approaches

Keywords

Diabetes, SMOTE Oversampling, Ensemble approach, soft voting

            A main goal is to build a ML based framework to predict the diabetes in future. For obtaining efficient results, we performed various procedures such as data cleaning, preprocessing etc. We handled imbalanced dataset by using Smote technique. We selected relevant features by employing various feature selection methods. We employed ML methods like RF, KNN, NB, SVM, GB, DT and LR. Then we integrated all ML methods by utilizing voting classifier.

3. Predictive Machine Learning Techniques for Diabetes Detection: An Analytical Comparison

Keywords

Diabetes prediction, classifiers, logistics regression, naive bayes, kstar

            In our research, to forecast the diabetes affected person, we utilized different ML methodologies like Naive Bayes, Logistic Regression, KStar, and Random Forest through the utilization of K fold cross-validation. A comparative analysis is performed in our study in terms of different metrics. As a consequence, Random Forest method provides greater efficiency. We conclude that, our suggested approach helps to diagnosis diabetes at its early stage.

4. Early detection of Diabetes using Machine Learning Techniques

Keywords

Artificial Intelligence, Disease Detection, Healthcare

            For the early forecasting of diabetes, we employed several ML methods including Logistic Regression (LgR), k-Nearest Neighbors (k-NN), Support Vector Machine (SVM), Random Forests (RF), XGBoost, and LightGBM. To find out the optimal method, we evaluated the efficiency of all these methods. In that, LightGBM offers better end results. At last, we proved that, our recommended system is more efficient than other previous researches.

5. An Advance Approach for Diabetes Detection by Implementing Machine Learning Algorithms

Keywords

Data mining, accuracy, F-1 score, robustness, Extra tree Classifier

            A concept of data mining and ML techniques are employed in our study for diabetes prediction. We examined the efficiency and severity of utilized methods by considering correlating accuracy and F-1 rankings. Here we performed comparative study for these methods. At last, Extra Tree Classifier performs better than SVM.  Our research aims to develop an innovative approach to assist the healthcare professionals for early diabetes diagnosis.

6. Detection and Classification of Type 2 Diabetes using Machine Learning Techniques

Keywords

Convolutional Neural Network (CNN), GLCM (Gray Level Co-occurrence Matrix), Diabetic Retinopathy (DR)

            A main concentration of our work is to forecast the diabetic retinopathy through the utilization of different ML methods such as CNN, GLCM, and RF. We examined the effectiveness of the employed methods by considering several metrics. From the analysis, we conclude that, CNN method provides highest outcomes. We also build a website by using Django to categorize the diabetic retinopathy into various types.

7. Comparative Approach for Early Diabetes Detection with Machine Learning

Keywords

Algorithms, Classification, Gaussian Naive Bayes, Support Vector Classifier, Decision Tree Classifier, K-Nearest Neighbors

To predict the diabetes by employing several ML methods is an ultimate goal of our study. We detected the optimal method by evaluating various techniques including Support Vector Classifier, Gaussian Naive Bayes, Random Forest, Decision Tree Classifier, Logistic Regression, Extra Tree Classifier, K-Nearest Neighbors, and XGBoost. Results show that, Extra Tree Classifier achieved greater efficiency than other methods.

8. Machine Learning Based Diabetes Detection Model for False Negative Reduction

Keywords

Pre-processed data, SMOTE, Balance data, Features selection

            We recommended an efficient ML framework in our research for anticipating diabetes disease. We employed many ML techniques such as LnR, LR, KNN, NB, RF, SVM, and DT. We preprocessed the data by eliminating null values and performing data standardization, normalization, and data labeling. We also overcome the problem of unbalanced data by utilizing SMOTE technique. From the analysis, we state that, RF method generates efficient end results.

9. Bio-Inspired Machine Learning Approach to Type 2 Diabetes Detection

Keywords

Bio-inspired, metaheuristic, chronic disease, type 2 diabetes, detection, cuttlefish

Enhancing the performance of previous techniques in the type 2 diabetes identification is the major objective of our article. After preprocessing the data, we selected relevant features by employing a bio-inspired metaheuristic method named cuttlefish. We compared the efficiency of our suggested method with other bio-inspired metaheuristic-based feature selection method denoted genetic method. As a consequence, cuttlefish method outperformed the other method.

10. Machine Learning Algorithms and Grid Search Cross Validation: A Novel Approach for Diabetes Detection

Keywords

            Several ML based techniques are examined in our study to forecast the diabetes diseases. We also carried out our approach with ML methods through the utilization of grid search cross-validation. We evaluated the performance of our recommended framework by using the specified PIMA diabetes dataset. Results show that, out of all utilized methods, Random Forest offers greater outcomes and the employment of grid search cross-validation also enhance the accuracy.