Customer Opinion Analysis and Topic Extraction Using Machine Learning

In today’s competitive market landscape, understanding customer sentiments through online reviews is imperative for businesses. However, the unstructured nature of this data poses challenges for traditional supervised learning methods. This paper proposes a methodology integrating semi-supervised learning and clustering techniques to analyze software reviews. Through extensive preprocessing and feature extraction, including Word2Vec embeddings and keyword extraction, the approach achieves nuanced language understanding. Clustering algorithms like HDBSCAN partition the data into meaningful clusters, facilitating insightful classification. Hyperparameter tuning ensures optimal performance, with findings indicating the effectiveness of the approach in discerning product-specific opinions. Evaluation metrics confirm the quality of obtained clusters, revealing distinct sentiments towards various software products. This research contributes to enhancing customer understanding and decision-making in software development and marketing strategies.