I have a task which is time series forecasting with two columns 
to predict Number_of column. so I wonder what is the approach to deal with these two time series to predict Number_of column.
This seems to require a model for a univariate time series. The Date-column is simply giving the temporal ordering of the counts. The only specialty here is that the data represents counts of small magnitude. Thus the data is clearly not continuous. On Rob Hyndman's website I found something that might be useful for modelling count data: https://otexts.com/fpp3/counts.html