Diabetes-Prediction

Dataset -

The dataset consist of 8 attributes and 1 output variable as follows:
1.'Pregnancies'- No of pregnancies the person had
2.'Glucose'- Glucose level of the person
3.'BloodPressure'- Bp of the person
4.'SkinThickness'-Thickeness of skin
5.'Insulin' - Insulin level of the person
6.'BMI'- Represents the body mass index of a person
7.'DiabetesPedigreeFunction'- data on diabetes mellitus history in relatives and the genetic relationship of those relatives to the patient
8.'Age'- Age of the person
9.'Outcome'- whether the person is suffering from diabetes (1) or not (0)

Some Insights in the dataset

Dataset had no missing values
Outliers were handled by visualizing them using box plot
Data was skewed so log tranformation was applied to the data

Steps

1. Data was visualized using seaborn and all the outliers and skewed data was handled

2. After that Correlation between attributes was calculated

3. A simple random forest classifier was used to determine if a person is suffering from diabetes or not

4. A RandomizedSearchCV was used to optimize hyperparameters of RandomForest classifier

Results

RandomForest Classifier - 95.33 %
RandomForest Classifier (Hpyertuned) - 97.98 %

Requirements

Python 3.*

Pandas

Matplotlib

Seaborn

sklearn

Links