The data quality of fiber optic networks can affect the communication quality of users. When abnormal data is included, it will significantly reduce communication efficiency and quality. To this end, a multi-channel fiber optic network big data anomaly detection algorithm based on high-dimensional spatial clustering is proposed. Obtain abnormal data features based on historical data feature density indicators; Using high-dimensional Gaussian mixture clustering algorithm to map data features from low dimensional space to high-dimensional space, in order to reduce computational difficulty, using kernel mapping to convert high-dimensional inner product calculation into low dimensional data kernel calculation; Finally, the HGMM algorithm is used to obtain the detection time series at different times, and the final detection result is output after fusion. The experimental results show that the proposed method can achieve a feedback rate of up to 98%, and the detected abnormal data types are completely consistent with the actual results, ensuring that the fiber optic network is not affected by abnormal data.