Design of K-Nearest Neighbor Algorithm For Classification of Credit Loan Eligibility At Senarak Dana Purwakarta Cooperative

Abstract

Semarak Dana Cooperative is a savings and loan cooperative located in Purwakarta that has experience lending money to its members as many as 162 money lending transactions. However, there are 26 instances of bad debts. To avoid bad debts, the cooperative needs to classify loans to its members. Classification is done by using the K-Nearest Neighbor (KNN) method based on the attributes of employment, income, age, credit amount, term, and collateral value. Data taken from as many as 162 members are sorted into 2 parts, namely 149 transactions used as training data and 13 transactions used as testing data. In addition, the data is also sorted into two classes, namely 136 current classes and 26 bad classes. The KNN process consists of four stages. First, determine the parameter K nearest neighbor distance. The second stage is to calculate the distance between testing data and training data using Euclidean distance. The third stage sorts the distance data that has been calculated using selection sort in order from the smallest to the largest value of K. The fourth stage calculates the largest number of classes for the largest number of classes set as the classification result class. Implementation using Borland Delphi and Mysql database. The research method was used by applying the Waterfall method. The Waterfall method used is composed of analysis, design, coding, and testing. System design using Unified Modeling Language (UML) by describing use case diagram, activity diagram, and class diagram. Based on the confusion matrix of the KNN classification process, the percentage of accuracy is 77%, precision is 88%, and recall is 78%. These results can be said that the results obtained are quite good, which exceeds 70%.