Machine learning based diabetes prediction and development of smart web application

Authors:
  • NANazin Ahmed
    Nazin Ahmed
  • RARayhan Ahammed
    Rayhan Ahammed
  • MIManowarul Islam
    Manowarul Islam
  • AUAshraf Uddin
    Ashraf Uddin
  • AAArnisha Akhter
    Arnisha Akhter
  • BPBikash Kumar Paul
    Bikash Kumar Paul

Top Cited Papers
Open Access
Abstract
Diabetes is a very common disease affecting individuals worldwide. Diabetes increases the risk of long-term complications including heart disease, and kidney failure among others. People might live longer and lead healthier lives if this disease is detected early. Different supervised machine learning models trained with appropriate datasets can aid in diagnosing the diabetes at the primary stage. The goal of this work is to find effective machine-learning-based classifier models for detecting diabetes in individuals utilizing clinical data. The machine learning algorithms to be trained with several datasets in this article include Decision tree (DT), Naive Bayes (NB), k-nearest neighbor (KNN), Random Forest (RF), Gradient Boosting (GB), Logistic Regression (LR) and Support Vector Machine (SVM). We have applied efficient pre-processing techniques including label-encoding and normalization that improve the accuracy of the models. Further, using various feature selection approaches, we have identified and prioritized a number of risk factors. Extensive experiments have been conducted to analyze the performance of the model using two different datasets Our model is compared with some recent study and the results show that the proposed model can provide better accuracy of 2.71% to 13.13% depending on the dataset and the adopted ML algorithm. Finally, a machine learning algorithm showing the highest accuracy is selected for further development. Next, we integrate this model in a web application using python flask web development framework. The results of this study suggest that an appropriate preprocessing pipeline on clinical data and applying ML-based classification may predict diabetes accurately and efficiently.

This publication has 16 references indexed in Scilit: