A study on evaluating and forecasting the demand for tourism accommodation facilities in Jinan City was conducted. Data was crawled from platforms such as Ctrip, GoWhere.com, and Meituan using Python crawler technology. The LDA topic model was applied to uncover the latent thematic structure within the text data, enabling effective classification of accommodation place labels. Machine learning methods, including XGBoost, CatBoost, LightGBM, and RandomForest, were used to screen the core indicators that affect the ratings of lodging places. An evaluation system was constructed by combining the hierarchical analysis method (AHP) and the entropy weighting method to determine weights. Visual analysis was performed to reveal differences in reception capacity across various themes, accommodation types and areas, as well as proximity to transportation, business districts, schools and scenic spots. Suggestions for optimization are proposed based on the hotel’s theme, rating distribution, geographic location, review count and price, aiming to enhance service quality and supervision, thereby increasing reception capacity and user satisfaction.