时间序列分析，AR模型，用于流数据预测与滤波

输入参数：y为原始数据矩阵，p为AR模型的阶数，la为自回归模型的遗忘系数

输出参数：预测值，置信区间，离群点等-Time series analysis, AR model for prediction and filtering data stream input parameters: y original data matrix, p is the order of the AR model, la self-forgetting coefficient regression model output parameters: predictive value, confidence intervals, outlier points