基于文本矢量特征的电影评分预测模型

Film Rating Prediction Model Based on Text Vector Features

电影评分是观众选择电影消费的一个重要依据。针对当前电影评分预测模型涵盖电影信息少、算法较为单一、预测准确性不高等问题,本文融合了电影特征信息与电影文本信息,提出了一种基于文本矢量特征的电影评分预测模型。首先,基于Word2Vec模型对分词后的电影文本进行向量化处理;然后,通过TF-IDF算法给予每个词向量对应的权重,生成文本矢量特征信息;最后,文本矢量特征信息和电影特征信息一起输入到LSTM (Long Short Term Memory networks,长短期记忆网络)模型进行评分预测。实验结果表明,对比单一的机器学习模型以及电影特征信息模型,该模型的准确率有明显的提高,可以有效地预测出电影的评分。

Film rating is an important reference for film consumption.The current film rating model has to face issues like insufficient film information,limited algorithm and low prediction accuracy.Therefore,this paper combines film feature information and film text information,provide a film rating prediction model based on text vector features.Firstly,vectoring the segmented film text based on the Word2Vec model.Secondly,using TF-IDF algorithm to match each word embedding with a corresponding weight and generating text vector feature information.Lastly,inputting text vector feature along with film feature information into LSTM(Long Short Term Memory networks)in order to generate predictions.In contrast of single machine learning model and film feature information model,experimental results show that the accuracy of this model is significantly improved and film rating can be effectively predicted.