A brand’s official Weibo contains a large number of comments that reflect people’s emotional responses to the brand’s products, services and other aspects. Therefore, negative emotional information can receive much attention and spread widely, so it is very important to manage and control this effectively. This paper uses sentiment analysis to establish a sentiment classification model of brand affect based on the official Huawei Weibo comments. First, the official Huawei Weibo comments are obtained and word2vec is used to preprocess and characterize the comment text. Second, three machine learning algorithms are used to learn sentiment classification for brand affect, namely support vector machines, random forests, and deep belief networks. Through the comparison of the resulting classification accuracy from the experimental results, the best model is selected and the brand’s negative emotional comments are obtained. Third, the word frequencies of these negative comments are obtained and the monthly trends of the proportion of high-frequency words and negative emotional comments are calculated. At the same time, the existing problems are analyzed and corresponding countermeasures and suggestions are proposed in order to solve this problem for an enterprise, so that the company’s brand strategies are adjusted in a timely and reasonable way.