Diabetes Prediction using K-Nearest Neighbors algorithm Without KNN Library

Overview :

This project aims to develop a machine learning model to predict diabetes using the K-Nearest Neighbors (KNN) algorithm without KNN library. The model utilizes patient data such as glucose level, blood pressure, insulin level, etc., to predict the likelihood of a person having diabetes.

Requirements :

• numpy
• pandas
• matplotlib
• seaborn
• scikit-learn
• scipy

Dataset :

Dataset Source : Kaggle

Attribute Name	Attribute Name
gender	Age
hypertension	heart_disease
smoking_history	bmi
HbA1c_level	blood_glucose_level
diabetes

Methodology:

1. Data Preprocessing: Cleaning the dataset, handling missing values, Anomaly and Outlier, and encoding categorical & nominal.
2. Feature Selection: Identifying relevant features that contribute significantly to the prediction of CKD.
3. Model Training: Implementing the K-NN algorithm.
4. Model Evaluation: Assessing the performance of the model using appropriate metrics such as accuracy, precision, recall, and F1-score.

Preview

Dataset Preview

Correlation Matrix

Accuracy Result

Usage

1. Clone the repository :

git clone https://github.com/DikkiKartajaya/DiabetesPrediction_KNN.git

2. Install the required dependencies :

pip install -r requirement.txt

3. Run the Jupyter notebook DiabetesPrediction_KNN.ipynb to train and evaluate the KNN model.

Contribution :

Contributions to the project are welcome! If you have any suggestions for improvement, feature requests, or bug reports, please feel free to open an issue or submit a pull request.