Processing timeseries problems using Regression

This is a simple framework to process multi-variables timeseries dataset using regression.

Most time series analysis methods focus on single variable data. It's simple to understand
and work with such data. But sometimes our time series dataset may containe multi-varibles.
For example, in marketing analysis, profit of a day may not only be decided by the number
of customers, but also depend on campaign, CM and so on.
It is harder to model such problems and often many of the classical methods do not perform
well.

Since regression methods are good at processing multivarible, we can simply turn our timeseries
dataset into training dataset for regression by exluding time columns. By doing that, we need
to make sure if data at each time point are independent. In another word, our dataset are not
affected by past data.

Restrictions

1.In general when using regression methods to make a prediction at a certain time point, data for
independent variables at that time point must be avaliable.
be avaliable to make predicitons.
2.Data at each time point must be independent.
3.Each independent variable must be unrelated. In many cases multi-variables regression analysis
fails to clarify the correlations between the variables.

• For detail about 2. and 3., see "Assumption of Independence".
  Here is a good explaination

How to use

See example.ipynb