Detecting Diabetics using Machine Learning

Inspiration

In India around 11.4% of people are diabetic and almost 15.3% are pre-diabetic and its still raising enormously due to improper diet and improper nutrients intake. So, this model will help the patient to put in there health details such as BP, alcohol consumption, smoking habbits and predicts whether the patient is diabetic (2) or pre-diabetic (1), normal (0).

What it does

By using Machine Learning Models such as Logistic Regression, Random Forest, XGBoost this model will classify the patient into different groups such as diabetic, pre-diabetic or normal patient.

How we built it

First i have done the data pre-processing where in all the null values and noise data was either removed or replace with the respective data. Then performed EDA for finding the correlation and figuring out which model fit for such kind of problem. Then picked up different models according to the EDA and trained the data with 80% of data and tested it on 20% of unseen data. Then evaluated the performance of the model by using the accuracy. Figured out XGBoost performed well with atmost of 84.87% accuracy.

Challenges we ran into

The main challenge which i faced is finding the appropriate dataset. This is a big hassel. Later i found this dataset in a UCI repository.

Accomplishments that we're proud of

I could able to make a machine learning model that performs with around 85% accuracy.

What we learned

I have learned about how to train and test the machine learning models and figured out how to optimize the accuracy of the model by dimensionality reduction.

What's next for Detecting Diabetics using Machine Learning

This model can be furthur improved by making a flask backend by which this can be given a user interface which allows any patients to get insights on their sugar levels.

Built With

• logistic-regression
• machine-learning
• python
• random-forest
• xgboost