Project Overview

Diabetes mellitus is an extremely life-threatening disease because it contributes to other lethal diseases, i.e., heart, kidney, and nerve damage, according to J Healthc Eng’s publication [1]. In this project, machine learning based approach has been proposed for the classification, early-stage identification and prediction of diabetes. These approaches and datasets are solely serve for practical application target in Cambodia. Slight training technique will be utilize during the implementation reflected on citizen health and lifestyle.

Model Selection & Techniques:

• Diabetes Classification: Random Forest (RF), multilayer perceptron (MLP), and Logistic Regression (LR)
• Predictive Analysis: Long Short-term memory (LSTM), Moving Average (MA), Linear Regression (LR). Consider RNN or more advance technique like GRU
• Experimental Evaluation Benchmark: PIMA Indian Diabetes Dataset accuracy, precision, recall
• Observation Outperformance (MLP):
• Feature Engineering
  • Extract relevant features from the dataset such as age, BMI, family history, blood pressure, etc
  • Perform feature selection techniques like PCA to reduce dimensionality and improve model performance

Training & Evaluation

• Training Technique
  • Cross-validation to ensure the model’s robustness
  • Hyperparameter tuning to optimize model performnace
• Benchmarking
  • Primary using PIMA but also test models on locally collected datasets
  • Compare the performance of different models (RF, MLP, LR) using metrics like AUC-ROC, F1 Score, etc
• Outperformance observation

Motive

Type 2 diabetes, the incurable nature of T2D along it its chronicity and silent progression, requires the condition to be diagnosed as early as possible and managed properly and promptly on a regular basis by patients, caregivers, and health care professionals to prevent or delay complications.

• Total adult population: 10,169,200
• 4.9% prevalence
• Total case (adult): 596,000 |