Stack Overflow serves as a widely-used, community-driven platform where developers seek assistance with programming-related issues. While the platform allows users to post questions and receive multiple answers, a significant portion of these questions do not culminate in an accepted solution. This lack of a clearly identified best answer often results in confusion for both the original poster and future visitors, as well as increased time spent navigating through numerous responses. To address this challenge, we present a method for automatically identifying the most promising answer among unaccepted ones. Our approach involves the application of text mining techniques to extract 13 informative features from a large dataset comprising 15,464 questions, 37,275 answers, and 72,025 comments. These features capture various textual, structural, and user-related aspects of the posts. The extracted data are then used to train machine learning models aimed at predicting the answer most likely to be accepted. The study focuses solely on English-language content available on Stack Overflow. The proposed method demonstrates promising performance, achieving an overall accuracy of 71% and an F1 score of 70%. These results suggest that automated answer recommendation can significantly enhance the user experience by reducing ambiguity and improving the efficiency of information retrieval on Q&A platforms.