With the explosive growth of online video content, web dramas have become an important component of entertainment consumption for a large audience. This article constructs a prediction model that combines big data processing technology and Support Vector Machine (SVM) algorithm. This model utilizes rich historical data features such as drama genre, cast, production costs, and promotional work to predict future audience satisfaction with online dramas. By optimizing model parameters and feature selection, analysis of monthly audience satisfaction data over the past year revealed that different types of dramas have a significant impact on audience satisfaction, with comedy genres consistently scoring above 85 points on average per month. The average monthly satisfaction score for ancient costume dramas is about 82 points. In contrast, the average monthly satisfaction score for thriller and suspense dramas is about 78 points. This indicates that there are different evaluations of drama due to personal preferences. War-themed TV dramas, due to their more serious themes, have a relatively stable audience, with an average monthly satisfaction score of about 75 points. This study constructs a predictive model that can effectively improve prediction accuracy and provide scientific decision support for film and television producers, investors, and platform operators. Conversely, this can help them better grasp market trends and create high-quality works that better meet audience expectations.