Journal is one of the media used for the publication of scientific work in the form of the latest research supported by strong, relevant and comprehensive evidence to prove the validity of the research. Research results published in a journal are often used as references in other studies as a development effort from previous research. In referring to scientific work, the public can utilize journal ranking sites to find the best quality journals. SCImago Journal Rank (SJR) is one of the reputable journal ranking sites that are integrated with the Scopus database. However, there is an inequality between the value of the SJR indicator and its quartile label in several journals. One solution that can be proposed to correct inequality that occurs in journal ranking data is the classification method. This study uses the K-Nearest Neighbor (K-NN) algorithm as a classification method and K-fold Cross-Validation as a validation method. The classification process is carried out in nine scenarios using 2-fold to 10-fold cross-validation. Each scenario gets 25 classification results with 1 to 25 nearest neighbors. The goal is to get the best classification performance based on the nearest neighbor parameters and the number of folds used. The best classification performance is obtained in the fifth scenario using 6-fold cross-validation and 16 nearest neighbors. Even so, the best average performance achieved from all scenarios based on accuracy scores only reached 63%. This raises the assumption that the K-NN method is considered unable to produce an optimal performance that approaches the SJR classification system.
展开▼
