A urinary tract infection (UTI) is an infection of the bladder or kidneys. UTIs may be caused by germs due to poor hygiene and tubes placed in drain tubes, and occur in some elderly living in nursing homes. In clinical situations, most caregivers help residents with urination by using diapers or catheters. However, the use of both medical devices has a risk of infection and requires caregivers’ labor and time to change. If it is possible to predict the urinary volume in the bladder without attaching any sensors to the body, it will ease the burden on caregivers and prevent residents from contracting the infection by not using diapers and catheters. Therefore, we propose a method for predicting urinary volume in the bladder without attaching a sensor to the body. In this study, the prediction phase is divided into two steps: First, we predict urinary volume in the bladder based on a model that demonstrates how urine accumulates in the bladder considering the absorption spectrum of urine obtained immediately after urination. Second, we correct urinary volume predicted in the first step of the multi-task Gaussian process (MGP). We performed a series of experiments to evaluate the proposed method and calculated the error rate between the actual urinary volume and the urinary volume predicted by the proposed model at the time of urination. The mean error rate of the proposed method was 8.40%.