Customm KNN V/S Scikit-learn KNN

Built KNN from scratch using Pandas and Numpy

1. Data Import

• Two datasets are imported from CSV files: one for the main dataset (CO2_Emissions.csv) and the other for prediction (CO2_prediction.csv).

2. Data Preprocessing

• Initial data preprocessing tasks are executed, including dropping unnecessary columns and checking for missing values.

3. Feature Engineering

• The dataset is separated into numerical and categorical features for further processing.

4. Data Transformation

• Data transformation techniques are applied, including scaling numerical features using MinMaxScaler and label encoding for categorical features using LabelEncoder.

5. Data Concatenation

• The scaled numerical and encoded categorical features are concatenated into a single dataframe, preparing the data for modeling.

6. Train-Test Split

• The data is split into training and test sets to evaluate the performance of the machine learning models.

7. Model Implementation

• A K-Nearest Neighbors (KNN) regression model is implemented using both a custom KNN implementation and the scikit-learn KNN model.

8. Model Evaluation

• The models are trained and evaluated using different values of k (number of neighbors), and mean squared errors are calculated for both training and test data.

9. Performance Comparison

• The performance of the custom KNN implementation is compared with the scikit-learn KNN model by calculating the mean squared errors (MSE) for their predictions.

10. Conclusion

• Based on the MSE values, it is found that the custom KNN implementation has a lower MSE, indicating better performance compared to the scikit-learn KNN model.