Business Valuation with Machine learning

The aim of the article is to test hypothesis about applicability machine learning methods train models that allow accurately predict market capitalization an enterprise based on data contained in three main forms financial statements: Income statement, Balance sheet, and Cash flow statement . scientific novelty study lies proposal alternative approach actual finance problem — business valuation. conducted empirical allows us under consideration. We various using most popular ( LASSO , Elastic Net KNN Random Forest SVM, others). To determine best for assessing value a company, effectiveness different compared R 2 performance metric (86,7% GBDT ). Financial statements NYSE NASDAQ companies are used. also addresses interpretability trained models. important features identified their specific items have greatest impact capitalization. Three independent ways feature importance indicate significance information In particular, Comprehensive income was item accurate predictions. Robust variable normalization missing imputation highlighted. Finally, improving developed recommended achieve even higher accuracy forecasts. concludes can be applied as more accurate, unbiased, less costly company. Feature analysis used understand further explore creation process.