<p>In the process of electroslag remelting, the high-temperature electrical conductivity and viscosity of slag are crucial for the selection of slag. Due to the differences in measurement temperature, composition, and experimental equipment used by various researchers for the quinary fluorine-containing slag (CaF₂–Al₂O₃–SiO₂–CaO–MgO), different predictive formulas and even contradictory conclusions have been obtained. This study collects data on the high-temperature electrical conductivity and viscosity of the quinary slag. Based on the characteristics of the data, a normalization scheme is proposed. A Self-built Learning Model is developed, which not only achieves higher prediction accuracy than traditional machine learning models but also allows for the calculation of the electrical conductivity and viscosity coefficients of each component in any slag composition. Furthermore, using the Self-built Parsing Model, the interaction coefficients of electrical conductivity and viscosity between different components are determined. This enables the prediction of electrical conductivity and viscosity with errors comparable to those of machine learning models. Unlike the black-box nature of machine learning models, the parameters of this model are visualizable, allowing researchers to directly utilize the parameters provided in this study for further research.</p>