Determining The Parameter K of K-Nearest Neighbors (KNN) using Random Grid Search

The study aims to optimize the classification of reception of training participants using the K-Nearest Neighbors (KNN) algorithm by finding the optimal value for the K parameter through a combination of search methods, Random search and grid search. This approach is expected to address the problem of manuality in the selection process of training participants who are susceptible to human error. The grid search method is used to search for parameter values in sequence, while the random search methods are used for optimum value search efficiency. The results showed that random grid search obtained accuracy of 91% with K=5, using manhattan metric, and Mean Absolute Error (MAE) of 0.081. These findings contribute to improving the efficiency and accuracy of the classification process by automatically accepting training participants using machine learning approaches, with results relevant to the development of similar systems in the future.