Satisfaction measurement, which emerges in every sector today, is a very
important factor for many companies. In this study, it is aimed to reach the
highest accuracy rate with various machine learning algorithms by using the
data on Yemek Sepeti and variations of this data. The accuracy values of each
algorithm were calculated together with the various natural language processing
methods used. While calculating these accuracy values, the parameters of the
algorithms used were tried to be optimized. The models trained in this study on
labeled data can be used on unlabeled data and can give companies an idea in
measuring customer satisfaction. It was observed that 3 different natural
language processing methods applied resulted in approximately 5% accuracy
increase in most of the developed models.

使用 Yemek Sepeti 的数据和其变体，采用多种机器学习算法，结合多种自然语言处理方法，旨在达到最高的准确率，并优化算法的参数，训练得到的模型可用于无标签数据的客户满意度测量。本研究中应用的三种自然语言处理方法能够使大部分模型的准确率提高约 5%。