In actual industrial sites, verifying the framework for cable manipulation is crucial. Therefore, it is necessary to simulate the deformation of the cable to predict its behavior accurately. By simulating the behavior in advance, it is possible to reduce the time and cost required for work. Although finite element analysis is used in various fields, the results may differ from the actual behavior depending on the method of defining the analysis model and analysis conditions. This paper aims to select appropriate indicators that can effectively cope with finite element analysis and experiments during cable winding work. We perform finite element analysis of the behavior of flexible cables and compare the analysis results with results from experiments. Despite some differences between the experimental and analysis outcomes, an indicator was developed through trial and error to align the two cases. Errors occurred during the experiments depending on the analysis and experimental conditions. To address this, weights were derived through optimization to update the cable analysis results. Additionally, deep learning was utilized to update the errors caused by material properties using the weights. This allowed for finite element analysis even when the exact physical properties of the material were unknown, ultimately improving the analysis performance.