Özet:

The aim of this study is to estimate the next year’s average share revenues of companies in the Stock Exchange Istanbul 30/100 and manufacturing industry companies in the Sustainability Index using year-end financial tables and economic indicators. Comparing the predictive performance of the methods used in the study under the distinctive characteristics of the methods constitutes the other purpose of the study. Method - For example, 13 companies in the 2010-2018 period took the annual financial statements and the annual average economic indicators and estimated the 2011-2019 annual average share revenues using the methods of artificial nerve networks (ANN), Classification and Regression Tree (CART) and the K Nearest Neighbor's Algorithm (KNN). The study used a 10-fold cross-verification method for optimal data distribution and parameters optimization for the purpose of determining the model that gives the highest estimated result in the given parameters range. The performance of the methods is compared to the ROC curve. Results - The analysis resulted in CART (%94.87) and ANN (%94.02) and KNN (%92.31) achieving the overall forecast and classification accuracy. Examples that provide negative return from 20% to 50% are estimated correctly at ANN (80,00%), CART (100%), KNN (100%) and at positive return estimates from 20% to 50% are estimated correctly at KNN (100%), ANN (96,55%), CART (96,55). The study found key variables that divide the data in the creation of the CART decision tree as the "Rate of Capital Profitability" and "Market Value /Defter Value". Discussions - All methods used in this study were classified and predicted by obtaining an overall prediction accuracy of more than 90%. Gaganis (2009) observed that more than 75% of the accuracy of the detection was a good result in the field of social sciences. According to Gaganis (2009) classification, the results of the ANN, CART and KNN analysis in this study can be evaluated very well.

Anahtar Kelimeler: