Performance Dataset for Multi-Component Sunscreens

This study systematically constructed a comprehensive multi-component sunscreen formulation dataset for machine learning model training and validation. The dataset comprises 600 formulations, including 30 blank base controls and 570 multi-component formulations. A combined approach of stratified sampling and Latin hypercube sampling was employed for experimental design, incorporating seven common ultraviolet filters: Octinoxate (OMC), Homosalate (HMS), Octocrylene (OCR), TiO₂, Avobenzone (AVB), Bis-Ethylhexyloxyphenol Methoxyphenyl Triazine (Tinosorb S), and ZnO. In all non-blank formulations, the total filter content was fixed at 18%, with the content of any single filter not exceeding 10% and all concentrations set as integer values.

The performance of the designed formulations was simulated using the DSM Sunscreen Optimizer simulator, generating data for five key performance indicators: Sun Protection Factor (SPF), UVA Protection Factor to SPF ratio (UVA-PF/SPF), Blue Light Protection Factor (Blue Light PF), UV Filter Efficiency, and Ecological Impact. The final constructed dataset contains 7,200 data points, comprehensively covering formulation compositions and their corresponding performance metrics. During the data preprocessing stage, missing value handling and feature standardization were performed. The dataset was then randomly split into an 80% training set and a 20% test set, providing a reliable data foundation for subsequent machine learning modeling.