Topic Modeling: Switching from implementation to XGBoost

[ghost]

Feature description:

Change the implementation of the current topic modeling to xgboost, since it obtained better results in new studies.

Research: https://drive.google.com/file/d/11CdWkqLgQ3f-1FNI1AvNwKEIkYO336Z_/view?usp=sharing and https://drive.google.com/file/d/16ifjhucCLsFbN3Y2PgZWfORj2-viepmg/view?usp=sharing

Need:
There was a great improvement in the model. Accuracy: 83%

Implementation:

• The texts were labeled with one of the chosen categories.
• Then, a study was carried out in the database, to remove the categories that appear in less than 90 texts.
• And then it was trained using XGBoost and BI-LSTM, and XGBoost achieved superior accuracy.