Research Article
Detecting Fake Reviews with Generative Adversarial Networks for Mobile Social Networks
Table 1
The comparison results of different models.
| Methods | Precision | Recall | F1-score | AUC |
| LR (TF-IDF) | 0.49 | 0.59 | 0.53 | 0.55 | XGBoost (word2vec + TextCNN) | 0.63 | 0.64 | 0.63 | 0.66 | AdaBoost (word2vec + TextCNN) | 0.70 | 0.59 | 0.64 | 0.62 | Random forest (word2vec + TextCNN) | 0.76 | 0.77 | 0.77 | 0.76 | SVM (TF-IDF) | 0.74 | 0.71 | 0.73 | 0.82 | KNN (TF-IDF) | 0.73 | 0.75 | 0.74 | 0.66 |
| TFM | 0.66 | 0.68 | 0.67 | 0.79 | Zhang et al.’s method [31] | 0.70 | 0.68 | 0.69 | 0.79 | Lyu et al.’s method [16] | 0.72 | 0.74 | 0.73 | 0.80 |
| Our model (word2vec + TextCNN + GANs) | 0.79 | 0.80 | 0.80 | 0.82 |
|
|