Recently, many researchers not only from industry but also from academia have interested in the art finance market. Although many people have studied growing art finance market, very few studies on the prediction model to evaluate the art price have been conducted. Therefore, in this study we have gathered a database of 12,105 paintings auctioned between 2009 and 2018 and have implemented decision tree-based machine learning algorithms (e.g., random forests, gradient boosting, XGboost) to develop prediction models for art prices and to improve the reliability for art experts’ estimates accurately. We have compared the prediction accuracy of the proposed approach based on root mean square error and mean absolute error. As a result of the analysis, we noticed that experts’ estimates from auction houses are more accurate only for high-priced artworks but overestimates low-priced artworks. On the other hand, the suggested prediction model’s accuracy considering decision tree-based model is better than the accuracy based on the parametric OLS model. Finally, we test the accuracy of prediction models considering expert evaluation to enhance the models’ predictive power. These results show that an integrated approach between expert appraisal systems and statistical models improves the prediction accuracy for artworks’ price.