The current data formats in the Internet of Things system are not unified,which makes it difficult to process,analyze and store a large amount of data.In this study,a new AI data fusion platform is established based on artificial pulse neural network to integrate and process multiple data sources.An improved pulse neural network is proposed,which combines the spatial information of BP neural network and converts the pulse information into time information through the change of membrane voltage.The output data of the output layer combines the spatial information and time information.The PALC model is established based on the multi-layer bidirectional LSTM network to extract the morphological and semantic features of the data,and extract and merge the entity information of the unstructured data on the platform.The experiment results show that the database of the data fusion platform in this study has a higher query rate per second,the highest QPS can reach 59 863,and the highest accuracy rate of information extraction is 99%.