Sentimental analysis for clothing dataset

Abstract

In recent years, sentiment analysis has been applied to various fields, including politics, business, education, etc. Simultaneously, with the development of science and quality of life, clothes have become an increasingly important part of people's daily lives, especially women. In this research, using Natural Language Processing, we interpret a dataset about clothing by trying different approaches, which include other models (CNN, RNN, and LSTM), different optimizers (Adam, RMSProp, and SGD), and different hyper-parameters (Epochs, Batch Sizes, and Learning Rate). Then, we analyze the influences of those approaches by all kinds of valuation standards. Ultimately, we improve our accuracy by 8.0 %.