Machine Learning–Based Financial Forecasting and Valuation Models for Investment Decision-Making and Capital Allocation Efficiency

Abstract

This study addresses a recurring enterprise problem: capital allocation decisions often rely on forecasts and valuation estimates that are inconsistent across tools and teams, leading to delayed approvals, mispriced projects, and inefficient resource deployment. The purpose of the study was to quantify how machine learning based forecasting and machine learning based valuation contribute to higher investment decision-making quality and improved capital allocation efficiency within cloud and enterprise decision contexts. A quantitative cross-sectional, case-based design was used, drawing on cloud and enterprise cases where ML dashboards and analytics were actively used in screening, budgeting, and portfolio governance. Data were collected using a structured 5-point Likert survey from N = 162 participants involved in the forecasting–valuation–allocation pipeline (analysts/associates 46.3%, managers/senior managers 32.1%, committee or strategic roles 21.6%). The key variables were ML Forecasting Effectiveness (FCAST), ML Valuation Effectiveness (VAL), Investment Decision-Making Quality (DMQ), and Capital Allocation Efficiency (CAE), with role and experience treated as controls; two additional indices were examined to clarify adoption conditions, Model Trust and Adoption Readiness (MTAI) and Forecasting–Valuation Alignment (FVAD). The analysis plan included reliability testing, descriptive statistics, Pearson correlations, multiple regression, and bootstrap mediation to test whether DMQ transmits the effects of FCAST and VAL to CAE. Reliability was strong (Cronbach’s alpha: FCAST = .88; VAL = .86; DMQ = .84; CAE = .87). Mean scores were above neutral (FCAST M = 3.94, SD = 0.61; VAL M = 3.88, SD = 0.64; DMQ M = 3.76, SD = 0.58; CAE M = 3.71, SD = 0.62), indicating generally favorable perceptions of ML support. Correlations were positive and significant (FCAST–DMQ r = .56; VAL–DMQ r = .52; DMQ–CAE r = .62; all p < .001). In regression, FCAST and VAL jointly predicted DMQ (R² = .41; FCAST β = .39, p < .001; VAL β = .31, p < .001), while CAE was explained by DMQ with additional direct effects (R² = .53; DMQ β = .45, p < .001; FCAST β = .18, p = .007; VAL β = .12, p = .049). Mediation results showed meaningful indirect effects via DMQ for forecasting (indirect = .18, 95% CI [ .10, .28]) and valuation (indirect = .14, 95% CI [ .07, .23]). Implementation implications are clear: enterprises should govern ML forecasting and valuation as an integrated decision system, strengthen model trust and adoption readiness (MTAI M = 3.82, SD = 0.55; 68.5% high readiness), and actively manage forecast–valuation alignment because misalignment is associated with lower allocation efficiency (FVAD–CAE r = −.34, p < .001).