In today’s rapidly developing Internet economy, people are increasingly preferring to consume and shop on self-media platforms. The change of marketing methods has brought huge business opportunities to businessmen. However, due to the problems of low threshold, wide mass clusters, few sales data samples and many channels in the sales of clothing on self-media platforms, forecasting sales of we-media platforms has become a huge headache which affects the survival of clothing enterprises. According to the characteristics of self-media apparel sales data, the article combines the literature research on gray prediction theory, selects and optimizes the multidimensional gray prediction GM (1, N) model. Besides, by means of data collection from target customers, correlation analysis, error test and model application, it reveals that the use of multidimensional gray prediction model has strong feasibility and high prediction value for the self-media platform apparel sales prediction, which provides apparel companies’ sales prediction a fresh idea.