车企舆情正负面识别与预测

Recognition and Prediction of Positive and Negative Opinions of Car Companies

摘要: 随着科技的不断进步，人们生活越来越好，车辆普及度逐渐提高，人们也越来越关注车辆带给他们的体验。而对于汽车企业而言，汽车安全直接关乎客户的生命安全，人们对于车企舆情的正负面也有着更高的关注度和敏感性，舆情处理难度只会更大。如果负面舆情不能及时处理，车企将面临着重大的舆论压力，而且事后进行处理时也会耗费大量的资源和财力。由于产品大多生产规模庞大，多方利益纠缠，车企的舆情系统往往比其他企业有更高的舆情要求，所以对于汽车企业而言，舆情的识别与预测起着很重要的作用。本文通过建立朴素贝叶斯模型对车企舆情正负面进行识别与预测，在有效处理数据的基础上，利用给出的训练集数据建立模型，用测试集数据对模型的合理性和科学性进行评估验证。研究表明，本文所采取的车企舆情识别与预测模型准确度较为理想，可靠性较强，但是将舆情倾向重新定义后，模型精度得到了较大提高，对于负面舆情的识别精度有了较大提升，本模型可以用于实际生活中车企舆情的判断。最后本文提出展望，在训练模型时数据选取时应尽量使得各类样本的数据占比均衡，避免造成过度识别问题。

Abstract: With the continuous advancement of technology, people’s lives are getting better and better, the popularity of vehicles is gradually increasing, and people are paying more and more attention to the experience that vehicles bring to them. For auto companies, car safety is directly related to the lives of customers. People are more concerned about and sensitive to the positive and negative public opinions of auto companies, making it more difficult to deal with public opinions. If the negative public opinion cannot be dealt with in a timely manner, car companies will face significant public opinion pressure, and it will also consume a lot of resources and financial resources when dealing with it afterwards. Since most of the products are produced on a large scale and multi-party interests are entangled, the public opinion systems of auto companies often have higher public opinion requirements than other companies. Therefore, for auto companies, the identification and prediction of public opinion plays a very important role. The paper establishes a Naive Bayes Model to identify and predict the positive and negative public opinion of car companies. On the basis of effective data processing, this paper uses the given training set data to build the model, and uses the test set data to evaluate the rationality and scientificity of the model. Studies have shown that the accuracy and reliability of the public opinion recognition and prediction model for car companies adopted in this article is relatively satisfactory, but after redefining public opinion tendencies, the accuracy of the model has been greatly improved, and the accuracy of identifying negative public opinions has been greatly improved. This model can be used to judge the public opinion of car companies in real life. Finally, this article puts forward a prospect that when selecting data when training the model, we should try to balance the proportion of data of various samples to avoid over-identification problems.