Driven by such special needs as "individuation, fashion and beauty", the production model of the new retail enterprise is gradually moving towards multiple varieties and small batches, which makes the variety of ornaments and toys in the retail stores in the shopping malls more dazzling, and at the same time, makes inventory management in the retail industry more difficult. How to give an accurate demand forecast based on the historical sales data with complex levels and various categories at the regional level, small category level and even the store skc (single monochrome) level is a problem that most new retail enterprises need to focus on and think about. First of all, we study the impact of various related factors on the sales volume of the target skc during the four holidays of National Day, Double Eleven, Double Twelve and New Year's Day in 2018. We first reviewed the literature and data to identify the influencing factors. We believe that factors such as product characteristics, inventory information and holidays will affect the sales volume. Secondly, we test the correlation coefficient of Pearson and Spearman on the sales volume and each factor, and obtain the correlation strength between each factor and the price, we think that those with strong correlation have a greater impact on the sales volume. Finally, we get the specific impact of each factor on the sales volume through regression fitting between each relevant factor and the sales volume. Secondly, we need to forecast the sales volume of a given region's target subcategory every month for three months, give the average absolute percentage error between each and the forecast value, and forecast the top ten subcategories of sales volume at the same time, first, we analyze the data and adopt the neural network model to forecast, give the monthly forecast data, and consider the impact of relevant factors, so that the results are more accurate. Finally, the monthly forecast value of MAPE is given.