python回归及岭回归_Python中比例的回归-CSDN博客

python回归及岭回归

I frequently predict proportions (e.g., proportion of year during which a customer is active). This is a regression task because the dependent variables is a float, but the dependent variable is bound between the 0 and 1. Googling around, I had a hard time finding the a good way to model this situation, so I’ve written here what I think is the most straight forward solution.

我经常预测比例（例如，客户活跃的年份比例）。这是一项回归任务，因为因变量是一个浮点数，但因变量的范围是0到1。在谷歌搜索中，我很难找到一种模拟这种情况的好方法，因此我在这里写了什么我认为这是最直接的解决方案。

I am guessing there’s a better way to do this with MCMC, so please comment below if you know a better way.

我猜想MCMC会有更好的方法，因此，如果您知道更好的方法，请在下面评论。

Let’s get started by importing some libraries for making random data.

首先，导入一些用于生成随机数据的库。

Create random regression data.

创建随机回归数据。

Shrink down the dependent variable so it’s bound between 0 and 1.

缩小因变量，使其在0到1之间。

Make a quick plot to confirm that the data is bound between 0 and 1.

快速绘图以确认数据在0到1之间。

All the data here is fake which worries me, but beggars can’t be choosers and this is just a quick example.

这里的所有数据都是伪造的，令我担心的是，但乞be不能成为选择者，这只是一个简单的例子。

Below, I apply a plain GLM to the data. This is what you would expect if you treated this as a plain regression problem

下面，我对数据应用普通的GLM。如果将其视为简单的回归问题，这就是您所期望的

Here’s the actual values plotted (x-axis) against the predicted values (y-axis). The model does a decent job, but check out the values on the y-axis – the linear model predicts negative values!

这是相对于预测值（y轴）绘制的实际值（x轴）。该模型做得不错，但是请检查y轴上的值-线性模型会预测负值！

1
1

Obviously the linear model above isn’t correctly modeling this data since it’s guessing values that are impossible.

显然，上面的线性模型不能正确地对此数据建模，因为它猜测的是不可能的值。

I followed this tutorial which recommends using a GLM with a logit link and the binomial family. Checking out the statsmodels module reference, we can see the default link for the binomial family is logit.

我遵循了本教程，该教程建议使用带有Logit链接和二项式族的GLM。查看statsmodels模块参考，我们可以看到二项式族的默认链接是logit。

Below I apply a GLM with a logit link and the binomial family to the data.

在下面，我将具有Logit链接和二项式族的GLM应用于数据。

Here’s the actual data (x-axis) plotted against teh predicted data. You can see the fit is much better!

这是相对于预测数据绘制的实际数据（x轴）。您可以看到适合度更高！

1
1

CPython 3.6.3
IPython 6.1.0

numpy 1.13.3
matplotlib 2.0.2
sklearn 0.19.1
seaborn 0.8.0
statsmodels 0.8.0

compiler   : GCC 7.2.0
system     : Linux
release    : 4.13.0-38-generic
machine    : x86_64
processor  : x86_64
CPU cores  : 4
interpreter: 64bit
CPython 3.6.3
IPython 6.1.0

numpy 1.13.3
matplotlib 2.0.2
sklearn 0.19.1
seaborn 0.8.0
statsmodels 0.8.0

compiler   : GCC 7.2.0
system     : Linux
release    : 4.13.0-38-generic
machine    : x86_64
processor  : x86_64
CPU cores  : 4
interpreter: 64bit