Research of facebook prophet function effect for handwritten digit prediction

Artificial intelligence technologies are being used in medicine, engineering and industry 4.0 applications with the development of today's technologies. We can see that deep learning and pattern recognition models creates artificial intelligence science if we get the heart of this discipline. Especially in 1989 Yann LeCUN achieved a significant breakthrough in pattern recognition from a different view point with recognizing handwritten digits using Convolutional Neural Network(CNN) on MNIST dataset. In this study, it is aimed that all predicted weights of classical deep learning models on MNIST datasets are re-predicted with Facebook Prophet function which is effectively used in economic forecasts and mathematical prediction analysis and compared with classical Convolutional Neural Network weight values.Indeed this comparison which uses classical CNN model weights as training input and predictes with time serie model Prophet and compares prediction performances that will give us a different viewpoint in this discipline. As a result more classical CNN model predictions which has more powerful performances that Facebook Prophet function in layer and kernel numbers as are more difficult predict than other weights. It will be developed a new technic for specifying disentaglend representation with this weights.These weights can be seen as weakness of CNN or deep learning models. Finally with applying data poisoning to this weights also can be used as more easier method for predicting objects or images using time series as Prophet function.