基于SVM女性服装型号推荐方法研究

Study on SVM Based Women's Dress Size Recommendation

摘要：针对网络服装销售中由于服装尺寸退货问题，根据不同年龄段女性体型特征数据，提出了基于机器学习方法的女性服装型号推荐方法，并给出了女性体型判别及预测的大体流程。首先，随机选择300位年龄在18～50岁的女性体型数据作为研究数据；其次，提取身高、背长、臂长、肩宽、颈围、臀围、胸围、腰围作为预测特征集，并对特征采用信息增益方法得到增益指数；再次，采用SVM方法和RBF核函数训练多个模型；最后，采用投票方式选取最终所属类，进行服装型号推荐。最终分类器采用测试集测试，结果表明模型预测准确度达到98％以上，预测结果可靠。

In allusion to sales return problem due to clothing size in online clothing sales, this paper proposes women＇s dress size recommendation method based on machine learning method in accordance with body shape features of women in different age stages and offers a rough process to distinguish and predict women＇s body shape. Firstly, 300 women aged between 18 to 50 were randomly selected and their body shape data of served as the research data; secondly, the height, back length, arm length, shoulder breadth, neck circumference, hip circumference, chest circumference and waist circumferences were extracted as a predictive feature set, and the gain index was gained through adopting information gain method for the feature set; thirdly, SVM and RBF kernel function were used to train multiple models; finally, the final class was selected with voting method for clothing size recommendation. The final classifier was tested by the test set. The results show that the preduetion accuracy of the model can exceed 98% , and the prediction result is reliable.