思路：建立表格通过pandas的操作对数据进行整理

主要是提取出安预测大小排序的真实类型

然后通过假设筛选 设什么概率为判断为正例（如 预测为正例的概率为0.39为正，那概率小于他的就会为反例）这样我就可以得到预测的样本

def precision_score(y_true, y_predict):
    FP = np.sum((y_true == 1) & (y_predict ==0 ))  # FN
    TP = np.sum((y_true == 0) & (y_predict ==0))  # TP
    return TP / (TP + FP)
def recall_score(y_true, y_predict):
    FN = np.sum((y_true == 0) & (y_predict == 1))#FN
    TP = np.sum((y_true == 0) & (y_predict == 0))#TP
    return TP/(TP+FN)

def TPR(y_true, y_predict):
    FN = np.sum((y_true == 0) & (y_predict == 1))#FN
    TP = np.sum((y_true == 0) & (y_predict ==0))  # TP
    return TP / (TP + FN)
def FPR(y_true, y_predict):
    TN = np.sum((y_true == 1) & (y_predict == 1))#TN
    FP = np.sum((y_true == 1) & (y_predict == 0))#FP
    return FP/(FP+TN)

难得写了