Shallow Learning vs Deep Learning in Recommendation Systems

Recommendation systems have become ubiquitous in modern e-commerce platforms, social networks, and online media services. These systems aim to provide personalized recommendations to users based on their interests and preferences. In this chapter, we investigate the use of shallow and deep learning methods in recommendation systems. We compare different shallow and deep learning models and analyze their performance on a data set containing images of different objects obtained from the Sentinel-1 satellite. As a recommendation system in this application, plant, water, building, and land areas were preferred for objects and Batman City of Turkey was chosen as the study area. Ten samples from these objects were selected and recorded with the help of GPS. Then, image analyses of these selected objects were made with the help of Sentinel-1 satellite images from May 1 to June 1, 2021. These image analyses were performed using Sentinel-1 polarization band parameters (VV-VH). Later, data sets containing VV-VH band parameters were created. By combining these data sets, a single data set was obtained. By applying shallow learning and deep learning models to this data set, objects were tried to be detected. The results showed that these objects can be detected using Sentinel-1 satellite signal parameters and shallow learning and deep learning models. Finally, we compared shallow learning and deep learning results in the recommendation system.