Development of a comprehensive methodology for assessing the investment attractiveness of high-tech companies in the e-commerce sector based on factor analysis and mathematical modeling

In the context of high uncertainty and turbulence of the Russian stock market, scientifically based approaches to assessing the investment attractiveness of public companies are of particular importance, taking into account both fundamental operational indicators and a wide range of external macroeconomic, competitive, regulatory and geopolitical factors. This study is devoted to the development of an innovative methodology that integrates traditional methods of multiplicative and discounted analysis with modern machine learning algorithms and stochastic modeling, which allows us to move from point estimates to probabilistic intervals of fair value, adaptive to changing market regimes. The paper systematizes 50 key cost drivers, divided into endogenous (operational and financial metrics, logistics infrastructure, monetization) and exogenous (macroeconomic indicators, competitive environment, regulatory changes, force majeure), and using methods of gradient boosting, LASSO regression and Bayesian optimization, their weight ranges are determined, which allowed to build a multi-level hierarchy of influence on the market capitalization. The proposed system of eight interrelated mathematical formulas includes stochastic dynamics of key indicators (turnover, operating profit, customer base), nonlinear interactions between growth drivers, multiplicative adjustments to geopolitical and competitive risks (taking into account the GPR and EPU indices), corporate governance awards, ESG rating, technological sovereignty and intellectual property, and There is also discounting of projected free cash flows over the medium-term horizon. Dynamic calibration of the weights is carried out through a multifactorial regime classifier that takes into account market volatility, inflation differential, business activity and technical indicators, which ensures that the model adapts to different phases of the economic cycle. Practical verification based on historical data (2010-2026) showed high predictive accuracy (R2=0.93, RMSE≈135 rubles, MAPE=3.8%), and the final confidence interval of the fair value of shares obtained using the Bayesian approach demonstrated the practical applicability of the developed tool for substantiating investment decisions, risk management and scenario planning in conditions of high uncertainty.