Cross market consumer behavior intelligence through corporate multimodal financial sentiment computing and predictive recommendation systems

Market-context-aware recommendation is increasingly relevant to corporate finance and accounting because consumer behavior can provide forward-looking information about firm demand, revenue prospects, and market performance, while corporate financial information may in turn influence consumer attention and purchasing decisions. Effective prediction therefore requires jointly modeling historical interactions, temporally aligned financial and accounting-related signals, changing market states, and cross-market associations. Existing methods generally treat external-signal encoding, dynamic relation learning, and uncertainty estimation as separate objectives and rarely align corporate financial information and market signals with subsequent user behavior under a strict temporal protocol. This paper proposes Cross-Market Consumer Behavior Intelligence (CMCBI), which integrates behavior-aligned financial signal learning, dynamic cross-market relation construction, and uncertainty-aware decision making through explicit shared-state dependencies. In particular, multimodal signals derived from financial news, firm-level financial information, and market indicators are aligned with subsequent consumer behavior and reused for both graph node representations and signal-dependent relation estimation. The resulting graph-enhanced representation jointly supports consumer preference prediction and uncertainty estimation, enabling uncertainty-adjusted ranking while providing a behavioral perspective on financially relevant changes in firm and market conditions. Experiments center on MIND-FinMarket as the primary financial-news-aware setting, where explicit financial-news and market-indicator inputs provide a natural environment for examining the interaction between financial information and consumer behavior. Amazon-CrossMarket, TAOBAO-MM, and KuaiRand-27K provide complementary non-financial evaluations of cross-domain generalization, multimodal scalability, and exposure-shift robustness. CMCBI significantly outperforms several sequential and static-graph baselines, although its gains over the strongest dynamic-graph and multimodal baselines are modest and not always statistically significant. More consistent improvements are observed in financial signal–behavior alignment, uncertainty calibration, risk-aware selection, and robustness to missing modalities, textual noise, market volatility, and exposure shifts. These findings suggest that integrating corporate financial signals with consumer behavior modeling can provide a useful computational framework for applications connecting recommendation analytics with corporate financial forecasting, accounting information environments, and data-driven managerial decision making.