顾洪建　张帆　万甜甜　张衡

摘 要：为了在“互联网+大数据+人工智能+区块链+物联网”高度信息化的社会精准预测汽车销量，本文首先利用词图、维特比等算法对汽车评价内容进行分词操作来获取关键词语;其次利用朴素贝叶斯分类器的方法对分词的结果进行计算，获得每条评论内容的情感指数;再次利用Norton模型的三代产品模型结合情感指数来组成拟合模型，同时利用最小二乘原理估计拟合模型的参数;最后利用估计的参数结合某款汽车的评论数据以及每个季度的汽车销量来验证模型，验证结果的准确性高达91.29%。基于此模型，企业可进行车型的销量预测，为合理规划生产和战略布局提供参考和依据。

关键词：词图 维特比 情感指数 朴素贝叶斯 Norton模型 最小二乘法

Sentiment Classifier Combined With Norton Model to Predict Car Sales

Gu Hongjian，Zhang Fan，Wan Tiantian，Zhang Heng

Abstract：In order to accurately predict the sales of cars in a highly informatized society of "Internet + Big Data + Artificial Intelligence + Blockchain + Internet of Things"， this article first uses word graphs， Viterbi and other algorithms to segment the car evaluation content to obtain the keywords; secondly， the article uses the naive Bayes classifier method to calculate the result of word segmentation to obtain the sentiment index of each review content; thirdly the article uses the three-generation product model of the Norton model combined with the sentiment index to form a fitting model， while the principle of the square method is used to estimate the fifteen parameters of the fitting model; finally， the estimated parameters are combined with the review data of a certain car and the car sales of each quarter to verify the model; the accuracy of the verification results is as high as 91.29%. This model can basically meet the actual forecasting needs， and can provide reference and basis for the reasonable production planning of the enterprise.

Key words：word graph， Viterbi， sentiment index， naive Bayes， Norton model， least square method

1 引言

一直以来汽车都是我国国民经济重要的支柱产业，改革开放以来，我国汽车产业快速发展，技术水平稳步增强，现已成为世界较大的汽车市场。精准的预测汽车销量不但可以为汽车产业的营销提供有力支撑，而且还有利于管理、生产、采购、物流等计划流程的优化。此外，销量预测还可以在一定程度上为车企获得健康持续发展的源动力提供保证。从今年市场表现来看，在疫情最严重的2月，我国汽车行业的生产和销售基本处于停滞状态，成为拉低全国经济指标的最主要因素。我国工业增长值同比增长速度和汽车同比增速及日均产量，均受疫情影响出现断崖式的波动，因此精准预测汽车销量对国民经济健康有序发展具有一定的推动，对十四五规划起到决定性作用[1-2]。……