0 votes

I want to predict solds for the next year. I have dates and I parsed them and I filtered for the year 2013.

I want to predict for the year 2014. When I stay to create model learning, I have the error :

Training Failed… Sorry

ufunc subtract cannot use operands with types dtype('<M8[ns]') and dtype('float64')


asked by yenami

1 Answer

0 votes

You can extract date components (day of month, day of week, month, year,...) from a parsed date, and set these new columns as features. And, as a target, you keep the "sales" column.

After the model is trained (and deploy in the Flow), you will need to create a new dataset on which you can apply the model. This dataset can have the dates of next year.

You can learn from our webminars: http://www.dataiku.com/resources/webinars/

I hope that helps,
answered by
741 questions
762 answers
483 users