A Decision Tree Approach to Develop Price Prediction Rules for Korea Allowance Unit and Renewable Energy Certificate

Abstract

This study attempted to analyze and predict the market prices related to carbon credit and REC (Renewable Energy Certificate) in the age of Net Zero response to the climate crisis. Although the price volatility of carbon credit-related securities is less than that of stock prices, the volatility is still high on a monthly basis. KAU and REC transaction data were extracted from Korea Power Exchange (KPX). The data used for dependent variables in the analysis is the seven-year data from 2015 to 2022. Macroscopic variables such as interest rates, exchange rates, and international oil prices were used as independent variables to predict transaction prices. Through decision tree analysis, a model with excellent performance by algorithm and parameter combination was first selected. As a result, the random forest generally shows higher performance than C4.5. However, parameters were adjusted in order to utilize C4.5, which has strengths in schematization of decision trees compared to random forests, and a model with a predictive rate comparable to that of random forests was selected. This paper suggests a schematic decision tree of KAU and REC in the preliminary experiment and could derive some decison rules to get optimal decision. This study has contributions in the following points. Several independent variables that affect carbon credits and REC price fluctuations were found. From this, we could draw the influence relationship between the target variable and the main independent variables. In addition, we could draw schematic diagrams that derive major rules and patterns through the decision tree method, moreover, it was possible to conduct multi-dimensional analysis for price prediction rather than mathematical modeling techniques.