Abnormal Collocations, a significant aspect of the network Micro-language, challenge the traditional linguistic norms while embodying the creativity and vitality of internet culture. Utilizing a robust semantic knowledge base, this study conducts a systematic analysis of abnormal Collocations in online micro-language, identifying their formation mechanisms and manifestations. Based on this analysis, this paper proposes a targeted approach for the identification and handling of abnormal collocations. The experiments have shown that the construction of relevant knowledge bases and the classification processing of abnormal collocations can significantly improve the accuracy of identifying abnormal collocations. The research offers novel perspectives and insights into the field of network linguistics.