IoT Data Fusion and Algorithms in Smart Cities

The data fusion and algorithm of the Internet of Things in smart cities aims to solve the problems of decision lag and resource waste caused by inconsistent formats, noise interference, and insufficient real-time performance of multi-source heterogeneous sensor data. First, the raw data is cleaned and normalized, and the spatiotemporal feature encoder is used to extract the periodicity and spatial topological correlation of the time series. Then, the improved TuckER tensor fusion algorithm is used to map the multidimensional data of traffic cameras, geomagnetic sensors, and social media into a low-rank joint embedding space, and congestion prediction and traffic light optimization are achieved through a graph convolutional network and LSTM hybrid model. Under the real data set of the main urban area of Hangzhou, this method achieved a traffic flow prediction accuracy of 97.1%, the event response delay was reduced to 0.53 seconds, and the average road traffic efficiency for the entire period was 0.64TTI, which verified the significant role of cross-domain data fusion and spatiotemporal joint modeling in improving the efficiency of urban intelligent management, and provided a scalable technical path for dynamic resource allocation.