Firstly, apologies if this question is obvious, I am new to Time Series Forecasting & ML in general.

I have an application whereby I collect prices from betting exchanges on an interval. This data is fed to a db to be stored and, on an interval, I train my Time Series model on this data. Once I get the data initially, I manually encode it using 1/0 vector. For example:

Soccer     [1, 0, 0]
Basketball [0, 1, 0]
Golf       [0, 0, 1]

There are 3 variables of this type:

  1. event_type
  2. exchange
  3. market_type

I am following this tutorial which utilises VAR in TS Forecasting. However, the dataset shown is not using categorical variables so I am at a loss here.

Once I read my data into a dataframe & drop the timestamp, I am left with the following data structure:

enter image description here

The dtypes of the dataframe are:

enter image description here

When I try & invoke statsmodel's VAR with my training data, I get the error:

ValueError: Pandas data cast to numpy dtype of object. Check input data with np.asarray(data).

From a lengthy Google, it looks as though the best way to deal with this is via converting to string then int(?):

To me, this seems a bit redundant as isn't that defeating the whole point of dummy encoding?

How should I go about handling this data & feeding it into an Auto Regression model?

Also, as I'm pretty new to ML, could you please let me know if I am on the correct track or not with my model choice? I attempted types of Linear Regression but this data (once plotted on a graph) will follow a similar stationarity to that of a stock market plot.

Thanks in advance :)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.