predict-merchant-s-trend

finally, we will need to be able to :

1 classify merchants by: life cycle(just onboarded / old fellow) / industry /

predict merchants' next month TPV by their classes

create a DNN for every circumstance,

only need user to guide on which columns are the labels, default the last column is the label

they can select on the label instead of input it.

2 can tell which is a good trend of merchants

provide the trend of merchants

most likely to be a linear regression question

calculate regression and tell the coefficient of it then compair it with level average growth rate to see if it's better.

3 take out the outlier of the TPV/RRR/ASP for each merchant

mark if it's a good one or a bad one