Abstract: A sentiment analysis method of online healthcare reviews based on feature weighted word vector was proposed in view of the professional, diverse and less normative features of online healthcare reviews. The Word2vec method was used to construct the word vector model, and the sentiment word set was extracted to improve the sentiment lexicon in the field of healthcare service. The dependency between subject words and sentiment words was identified according to the syntactic relations. The expected cross entropy factor was introduced to establish a feature weighted word vector model to analyze the sentiment tendency of online healthcare reviews. The experimental results show that the accuracy, recall rate and <i<F</i<<sub<1</sub< value of the expanded healthcare service sentiment lexicon are higher than those of the basic sentiment lexicon. After the introduction of the expected cross entropy factor, the sentiment analysis method based on the feature weighted word vector shows better effect in the SVM classification, which reflects its good utility in the online healthcare reviews mining.