给定10000个用户和他们对10000个电影的评价，然后通过协同过滤或梯度下降算法，用训练集训练数据，预测出用户对未看的电影的评分，并与测试集对比验证预测结果的准确性(You can learn Chinese,and read the Chinese introduction.)

训练SVR模型做预测,可调整训练集和测试集比例及SVR参数，预测性能用MAP反映(The training SVR model can be used for prediction. The proportion of training set and test set and the parameters of SVR can be adjusted. The prediction performance can be reflected by MAP.)