Due to the limited accuracy of current information fusion methods, this study proposes an economic information fusion approach for the Internet of Things (IoT) based on genetic algorithms. The paper extracts entities and relationships related to events to construct event structures. Using a graph database enhances data relationships and connectivity, facilitating better representation and analysis. A region mining algorithm is employed to extract information from the complex transaction network formed by IoT economic data, improving depth of analysis. Varying numbers of cross individuals are selected to enhance population evolution efficiency. A new dispersion formula based on KL divergence is used to differentiate evidence, assigning varied weights accordingly. Subsequently, the evidence's information quantity determines final weight, modifying evidence credibility and yielding a weighted average body of evidence. Before applying the Dempster combination rule, the final weight adjusts the evidence body, followed by self-fusion to derive the ultimate fusion outcome. Experimental results demonstrate the reasonable fusion outcomes of experimental group A3, correctly identifying object1 with heightened accuracy.