Travel assistance devices and programs obtain a captured image that captures a surrounding environment of a mobile unit and obtain, as a map information image, a map image that three-dimensionally represents a map and that represents a same range as an image capturing range of the captured image from a same direction as an image capturing direction of the captured image. The devices and programs extract a risk factor based on learning data by inputting the captured image and the map information image to machine learning as input images with a plurality of channels. The risk factor is present in the surrounding environment of the mobile unit and is not included in the captured image.