This paper applies the random forest algorithm to classify the communication effect of the new media of Meihuahuas enterprise, based on the communication effect setting rules, so that the training samples of the communication effect evaluation are split according to the feature attributes and recursively generated a tree. By introducing cost-sensitive factors, the classification accuracy of effect evaluation is improved. Combining the cost-sensitive random forest algorithm, the communication effect evaluation index of Meihualu Enterprise’s new media is constructed. According to the results, the Meilu enterprise’s overall new media communication effect is excellent. The audience can be guided to consume to a certain extent by it. The score of the influence of the new media communication of Meilu enterprise on the audience’s behavior is 0.646. In evaluating the first-level indicators of the new media communication of Meilu enterprise, the scores of communication power, influence, and credibility are all greater than 0.1, and the scores of the three are 0.1143, 0.1746, and 0.1012, respectively.