Exploiting Behavioral Features to Detect Fake Reviews by Means of Contextual Features
Dissertation is available at https://arxiv.org/abs/2003.00807
- Java version 1.7.x
- common-lang3.jar
- commons-csv-1.4.jar
- javax.json.jar
- javax.json-api-1.0-sources.jar
- joda-time.jar
- jollyday.jar
- mysql-connector-java-5.1.43-bin.jar
- opencsv-4.0.jar
- poi-3.9.jar
- protobuf.jar
- sqlite-jdbc-3.7.2.jar
- stanford-corenlp-3.5.2.jar
- stanford-corenlp-3.5.2-models.jar
- xom.jar
the proprocessing code is written in Java. The downloaded datasets were in form of SQLite, that is why we initially statblish connection with database [com.yelp.database]. A customize engine [com.engine.*] is develop to extract features from text data. The code of overall number of feature extraction can be found in [com.yelp.rest.FeatureExtractor].
- Python verion 3.x.x
- SkLearn 0.18.+
- Pandas 0.20+
Pre-processing step will generate .CSV file. One can change the filename in given two files.
python random_forest_rest.py
python svm_res.py
A sample dataset is given in data folder