Financial Risk Estimation with KNN Classification Algorithm on Determined Financial Ratios

Abstract

Analyzing the financial situation of companies is very important today. Thanks to early analysis, companies can improve their financial situation and be saved from bankruptcy. Until now, many studies have been carried out on the financial analysis of companies. Data mining techniques are widely used in many different areas of the financial sector. In this study, data mining classification technique has been used to determine the financial risk status of companies. The data set has been created by using the balance sheets taken from the Kamu Aydınlatma Platformu (KAP) between 2013 and 2018 of different companies serving in the manufacturing industry. There are a total of 1027 records in the data set. On these records, the financial ratios determined for that year for each firm have been calculated and recorded in the database. The Springate model has been used to determine the financial success of the companies. Springate formula consists of 4 financial ratios. Springate score value is calculated by multiplying these ratios with certain coefficients. If the calculated score value is less than 0.862, the firm is considered unsuccessful, else the firm is considered successful. In this context, the Springate score has been calculated for each record and the companies have been labeled as successful or unsuccessful according to the score value. In this study, the KNN (K-Nearest Neighbor) algorithm has been used for classification. The KNN algorithm classifies the new element to be classified by looking at its proximity to its k neighbors. KNN algorithm needs learning data to classify unclassified data. In this study, cross validation method has been used for determining learning data and the data to be classified. According to the cross validation method, the data set is divided into n groups. For each group, the selected group creates the data to be classified. All other groups (n-1 groups) are used as learning data. Then the classification process is applied and the results are stored. After the classification process is completed for all groups, the statistical summary of the results is checked. In this study, the dataset was divided into 10 groups and the classification results have been evaluated. When the results have been examined, it has seen that the algorithm has make a successful classification with a rate of 88.42%. Thus, the financial risk estimation of the companies has made with the classification technique applied on the determined finance model.