#0599 · 方法
先立一个最笨的基线
做预测模型之前,先给出最笨的基线(比如永远猜平均值或上一期的值),报出它的误差。你的模型必须明显打赢它,否则就不值得用。
Before building any forecasting model, set up the dumbest baseline (always predict the mean, or last period's value) and report its error. Your model has to beat it clearly, or it isn't worth using.
小杜
已贴 1,040 张上墙共收到 0 个章