Web-Scraping of reviews from Trustpilot.com and the use of statistical and machine learning methods to analyze these reviews.
I used the BeautifulSoup library to scrape the reviews from the Trustpilot.com. I chosed bank service GoHenry for reviews analysing.
An exploratory analysis of the data was then carried out. I have reviewed the distribution of the ratings of the reviews. The presence of a linear relationship between the average monthly rating and the number of reviews per month was investigated.
Using NLP algorithms, the main features of positive and negative reviews were identified. Relevant conclusions were drawn about what influences the positive and negative assessment of the rating.
With the help of machine learning algorithms, classification models were built to predict the rating depending on the content of the review. I used Logistic Regression, SVM and Gradient Boosting models to predict the rating. Accuracy estimates for each of the models are presented in the notebook.