This project focuses on predicting whether a person has diabetes based on their medical measurements using the K-Nearest Neighbors (KNN) algorithm.

• File: diabetes.csv
• Rows: 768
• Features: 8 numeric features + 1 target column (Outcome)
• Target: Outcome → 0 = non-diabetic, 1 = diabetic

• Checked for missing or zero values in the dataset.
• Replaced invalid zeros with the mean or median of the respective column.

• Shuffled the dataset to remove any inherent order.
• Split the data into training set (80%) and test set (20%).

• Standardized features to have mean = 0 and standard deviation = 1 using StandardScaler.
• Scaling helps KNN perform better, since distances are sensitive to feature magnitudes.

• Calculated Euclidean distances between test samples and training samples.
• Selected the k = 5 nearest neighbors.
• Predicted the class by majority vote of the neighbors.

• Used KNeighborsClassifier for comparison.
• Applied standardization with StandardScaler.
• Verified that the scikit-learn model gives similar results to the manual implementation.

• Compared predictions to actual labels on the test set.
• Calculated accuracy and F1-score.
• Checked precision and recall for each class to see how well the model identifies diabetic vs non-diabetic cases.

• Manual KNN Accuracy: ~72%
• scikit-learn KNN Accuracy: ~76%
• Manual implementation helped in understanding the inner workings of KNN and the effect of scaling and distance calculation.