The article examines the development of algorithms for evaluating the efficiency of investment projects using the STEM approach and machine learning methods. It analyzes the need to shift from traditional valuation methods based on discounted cash flows (NPV, IRR, PI, DPP) to intelligent models that ensure adaptability and reproducibility of calculations. It is emphasized that the integration of scientific, engineering, technological, and mathematical components enables the formalization of evaluation processes and improves forecasting accuracy. The practical section presents modeling based on project data processed using XGBoost and LSTM algorithms, which demonstrated increased NPV forecasting accuracy and reduced computation time. The results confirm the effectiveness of the STEM-oriented approach in investment analysis.