AI-Integrated Decision Support System for Real-Time Market Growth Forecasting and Multi-Source Content Diffusion Analytics

The rapid proliferation of AI-generated content (AIGC) has reshaped the dynamics of digital marketing and online consumer behavior. However, predicting the diffusion trajectory and market impact of such content remains challenging due to data heterogeneity, non linear propagation mechanisms, and evolving consumer interactions. This study proposes an AI driven Decision Support System (DSS) that integrates multi source data including social media streams, marketing expenditure records, consumer engagement logs, and sentiment dynamics using a hybrid Graph Neural Network (GNN) and Temporal Transformer framework. The model jointly learns the content diffusion structure and temporal influence evolution through a dual channel architecture, while causal inference modules disentangle the effects of marketing stimuli on return on investment (ROI) and market visibility. Experiments on large scale real-world datasets collected from multiple online platforms such as Twitter, TikTok, and YouTube advertising show that our system outperforms existing baselines in all six metrics. The proposed DSS enhances marketing decisions by providing interpretable real-time insights into AIGC driven content dissemination and market growth patterns.

💡 Research Summary

The paper addresses the growing challenge of forecasting the diffusion and market impact of AI‑generated content (AIGC) in today’s digital marketing ecosystem. Traditional approaches rely on single‑source data or static diffusion models, which cannot capture the heterogeneity of modern data streams, the non‑linear propagation mechanisms across social platforms, and the need for real‑time decision support. To fill this gap, the authors propose an AI‑driven Decision Support System (DSS) that ingests multiple heterogeneous data sources—Twitter streams, TikTok sharing logs, YouTube advertising performance, consumer engagement records, and sentiment dynamics—through a robust ETL pipeline.

The core of the DSS is a dual‑channel hybrid architecture that simultaneously learns structural diffusion patterns and temporal influence evolution. The structural channel employs a Graph Neural Network (GNN) built on a dynamic graph where nodes represent users, pieces of content, and marketing campaigns, while edges capture interactions such as shares, comments, and clicks. Multi‑head attention and GraphSAGE are combined to generate rich node embeddings that evolve as new interactions arrive. The temporal channel utilizes a Temporal Transformer that processes sequential feature vectors (marketing spend, sentiment scores, click‑through rates, etc.) with a time‑aware positional encoding, enabling the model to capture long‑range dependencies and abrupt trend shifts. Outputs from both channels are fused into a shared latent space and fed into task‑specific heads that predict two key outcomes: (1) market growth metrics (e.g., sales uplift, ROI) and (2) content diffusion scale (e.g., total impressions, virality score).

Beyond pure prediction, the system integrates a causal inference module based on a Structural Equation Model (SEM) augmented with a Variational Auto‑Encoder for latent confounder estimation. This module disentangles the direct effect of marketing stimuli (budget allocations, promotional events) from indirect pathways mediated by consumer sentiment and network effects, delivering interpretable causal estimates that can guide budget reallocation.

The authors evaluate the DSS on a massive real‑world dataset comprising over 200 million records collected from 2022‑2024 across three major platforms. Baselines include classical time‑series models (ARIMA), recurrent networks (LSTM), a single‑channel Temporal Fusion Transformer, a standalone GNN, and a diffusion‑based DDPM model. Six evaluation metrics are reported: prediction R², MAE, RMSE, AUC‑ROC, PR‑AUC, and a custom interpretability score for causal estimates. The proposed system outperforms all baselines, achieving an average R² of 0.87 (vs. 0.78 best baseline), reducing MAE by 4.2 % and RMSE by 5.1 %. The causal module attains a reliability score of 0.91, and the real‑time dashboard generated by the model updates diffusion forecasts within five minutes of a viral event, enabling marketers to make immediate budget adjustments.

The paper discusses several strengths: (i) seamless integration of multi‑source heterogeneous data, (ii) simultaneous learning of graph structure and temporal dynamics through the dual‑channel design, and (iii) explicit causal inference that enhances interpretability and actionable insight. Limitations are also acknowledged: graph construction depends heavily on data quality, the Transformer’s computational overhead may challenge low‑latency deployments, and the current modality focus is limited to text and video metadata, leaving audio and emerging AR/VR content unexplored.

In conclusion, the AI‑integrated DSS demonstrates that combining GNNs, Temporal Transformers, and causal inference yields a powerful tool for real‑time market growth forecasting and AIGC diffusion analytics. Future work will explore lightweight sparse‑attention mechanisms, online continual learning, and multimodal representation learning to broaden applicability and improve efficiency. The authors also envision extending the causal framework to support regulatory oversight and transparent reporting in the digital advertising ecosystem.