Proceedings of the International scientific and practical conference ―Oxford International Science Forum‖ (April 17-19, 2026) / Publisher website: www.naukainfo.com. - Oxford, United Kingdom, 2026. - 367 p.

107 • to predict campaign outcomes at different time points; • to segment campaigns into analytically meaningful classes; • to detect anomalous or suspicious patterns; • to generate interpretable conclusions for campaign creators, investors, researchers, and platform managers [2, 4-6, 11-13]. Thus, CD mining should be considered an integrated task of heterogeneous data processing, pattern identification, predictive analytics, and decision support rather than a narrow classification problem. Discussion . This review shows that CD mining has evolved from a simple exploratory analysis of campaign success to a broader analytical field that includes multimodal data processing, semantic analysis, predictive modeling, and interpretive decision support [1, 3-4]. One of the key findings of the reviewed literature is that no single type of feature is sufficient to explain campaign results. Instead, meaningful analytical results arise from a combination of platform metrics, social signals, text attributes, and dynamic funding behavior [1-2, 4-6]. Another important finding of the review is that the quality of analytical findings depends largely on how data is collected and harmonized. As platforms differ in openness, architecture, terminology, and business logic, robust research should incorporate platform-specific preprocessing and carefully defined target variables [7- 11]. This is particularly relevant when combining data in reward-based, donation- based, and equity-based environments. The reviewed studies also indicate that the field is moving towards more sophisticated and understandable analytical processes. Along with forecasting accuracy, interpretability, reliability, and methodological transparency are becoming critical requirements. For this reason, modern CD analysis tools should not only generate estimates and forecasts, but also explain the signals and factors underlying these results [3-6]. Future research should focus on experimental validation of the considered analytical approaches on real crowdfunding datasets collected from different types of platforms. Particular attention should be paid to cross-platform data harmonization,