Fake-Review-Detection

Exploiting Behavioral Features to Detect Fake Reviews by Means of Contextual Features

Dissertation is available at https://arxiv.org/abs/2003.00807

Preprocessing Project Required Jars

• 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

About proprocessing code

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].

Classifier code in Python

• Python verion 3.x.x
  • SkLearn 0.18.+
  • Pandas 0.20+

Run Classifiers

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

Detail Study and Experimentation Results

Dataset

A sample dataset is given in data folder

author's email: [email protected]