The data features have a discrete distribution state in the network space, which leads to a relatively low reduction in the network space after eliminating data redundancy, affecting the effectiveness of redundancy elimination. Therefore, this study focuses on data centers and iteratively eliminates redundant data in optical interconnect network resources, taking into account the activity of data points. Firstly, based on the feature similarity of data resources in the data center's optical interconnection network, the partition boundaries of data resources are set, and the classification components are calculated. By dynamically searching for redundant features of resources, redundant data is separated from normal data to fundamentally ensure the effectiveness of redundancy elimination; Then, combined with the roughness function, analyze the discrete distribution state of redundant resource features in space and compress the redundant parameters; Finally, deep neural network algorithms are used to calculate the activity of redundant data points, filter out resources that exceed the activity threshold, and achieve iterative elimination of redundancy. The experimental results show that after applying this method, the maximum spatial reduction ratio of the network can reach 0.55, indicating that this method has significant effectiveness.