In this application, a predictive model is built for finding the price of an item from their characteristics, especially the description. The implementation of word vectors, optimized gradient descent, and long short-term memory are employed to tackle the problem. And the relatively low loss for both training and test set proves this is a reasonable approach.
