Quantitative assessment of the value of intangible cultural heritage art supported by multimodal machine learning

A scalable multimodal machine-learning (ML) approach for evaluating the cultural and aesthetic significance of fine-art paintings classified as intangible cultural heritage (ICH) is presented in the research. A hybrid Artificial Lizard Search–optimized Multi-Kernel Support Vector Machine with Long Short-Term Memory (ALS-MKSVM-LSTM) model is presented to forecast numerical valuation ratings because traditional expert-based assessments are frequently subjective and inconsistent. The ICH Fine Art Valuation dataset, which was assembled from open-source sources, had 2850 assigned artworks from which features were derived from image, textual, and historical data. A weighted ensemble of MKSVM and LSTM was utilized to merge visual, linguistic, and contextual features. ALS optimization was then employed to fine-tune feature selection and hyper-parameters for robust regression. Experiments on this dataset show that the suggested model outperforms the baselines of traditional CNN, RNN, and SVM with an accuracy of 97.0%. In evaluating the artistic and cultural worth of Chinese paper-cutting artworks, for instance, the model was able to identify high-value items by considering their historical significance and intricate designs. This illustrates the ways in which the system can assist museums, curators, and legislators make well-informed choices on conservation and exhibition. Overall, the suggested methodology provides a clear, effective, and expandable approach to quantitative heritage valuation, utilizing technology-driven research to support cultural identity preservation.