Evolving market conditions require power marketers to balance stability, customization, and efficiency. Strategic improvements in customer engagement and resource management now depend on advanced operational analytics and data-supported planning. The paper first sorts out the analysis subject of power marketing business, constructs a preliminary collection and integration of cross business system data with user code as the core identifier, and conducts data aggregation and quality evaluation around the theme analysis application requirements. The data is encoded and partitioned based on the scope and scale of the analysis topic, and map marketing related data into a set of business data modules according to the topic. Adopting the particle swarm algorithm with the introduction of cross attention mechanism to identify and extract data features, and designing a specialized method for analyzing electricity marketing themes based on neural network classification and prediction models. According to the data module division and the data requirements of the topic analysis algorithm, the matching scenarios between the analysis topic and the data-set are arranged to achieve automatic adaptation between the power marketing business topic and the data analysis algorithm template. Select multiple scenarios for electricity marketing business analysis, such as implementing preferential electricity prices, recovering electricity fees, and charging piles, to verify the applicability of the method. Compared to customized power marketing data analysis methods designed for specific thematic applications, this paper adopts a theme-application-dataset generative analysis approach based on cross-thematic fusion, which enhances the alignment between analytical results and user requirements, thereby improving power marketing data analysis capabilities in complex scenarios.