- Problem Statement
Detection of fake reviews out of a massive collection of reviews having various distinct categories like Home and Office, Sports, etc. with each review having a corresponding rating, label i.e. CG(Computer Generated Review) and OR(Original Review generated by humans) and review text.Main task is to detect whether a given review is fraudulent or not. If it is computer generated, it is considered fake otherwise not.
- Description: The generated fake reviews dataset, containing 20k fake reviews and 20k real product reviews. OR = Original reviews (presumably human created and authentic); CG = Computer-generated fake reviews.
- Numpy
- Pandas
- Matplotlib.pyplot
- Seaborn
- Warnings
- nltk
- nltk.corpus
- String
- sklearn.naive_bayes
- sklearn.feature_extraction
- sklearn.model_selection
- sklearn.ensemble
- sklearn.tree
- sklearn.linear_model
- sklearn.svc
- sklearn.neighbors
- Removing punctuation character
- Transforming text to lower case
- Eliminating stopwords
- Stemming
- Lemmatizing
- Removing digits
- CountVectorizer Bag of Words Transformer
- TFIDF(Term Frequency-Inverse Document Frequency) Transformer
- Logistic Regression
- K Nearest Neighbors
- Support Vector Classifier
- Decision Tree Classifier
- Random Forests Classifier
- Multinomial Naive Bayes
Support Vector Machines Classifier performed the most accurate predictions regarding the fake nature of reviews having a predictive accuracy of just over 88%, closely followed by Logistic Regression which had a prediction accuracy of a little more than 86%. Random Forests Classifier and Multinomial Naive Bayes algorithm predicted to a precision level of approximately 84%. However, the Decision Tree Classifier performed fake reviews prediction upto an accuracy of just over 73%. The worst performing algorithm was the K Nearest Neighbors algorithm which could only perform the predictions upto an accuracy level of nearly 58%.