User review analysis of dating apps based on text mining.

Area of Science:

• Information Technology
• Data Science
• Human-Computer Interaction

Background:

• Increased adoption of online dating apps, accelerated by the COVID-19 pandemic.
• Prevalence of negative user reviews for mainstream dating applications.
• Need to understand the root causes of user dissatisfaction.

Purpose of the Study:

• To investigate the primary reasons behind negative user reviews on online dating platforms.
• To develop and evaluate a machine learning model for classifying user reviews.
• To provide actionable insights for dating app operators to enhance services.

Main Methods:

• Application of topic modeling to identify key themes in negative user reviews.
• Development of a two-stage machine learning model for review classification.
• Utilized Principal Component Analysis (PCA) for data dimensionality reduction and XGBoost for classification after oversampling.

Main Results:

• Identified core issues driving negative feedback: charging mechanisms, fake accounts, subscription models, advertising, and matching algorithms.
• Demonstrated that PCA combined with XGBoost yields improved accuracy in classifying user reviews.
• Proposed specific improvement suggestions based on the identified issues.

Conclusions:

• User dissatisfaction with dating apps stems from specific functional and operational aspects.
• Machine learning techniques, particularly PCA and XGBoost, are effective for analyzing and classifying user feedback.
• Addressing identified issues can lead to improved user satisfaction and sustainable business models for dating apps.