How to define a custom factor using fundamental data of Morningstar dataset

define a new class with my factor's name

class my_factor(Custom Factor):
# Then I need my inputs and window-length
input =[]
window-length =

# Then I need to compute def compute(self, today, assets, out, close, shares): out[:] =

Probably, coding should be like this. But I am not sure where should I show it is fundamental data?
And what factors can I define? If anyone can show me some example with instructions that will help a lot.

Thank you for help.

Welcome! Hope this explanation helps.

Custom factors have dataset columns as their inputs. Examples of a dataset are USEquityPricing and Fundamentals. Examples of a dataset column are USEquityPricing.close and Fundamentals.cash_and_cash_equivalents. Really just attributes of a dataset. See the docs for a bit more explanation https://www.quantopian.com/help#importing-datasets .

So, to create a custom factor one needs to import the `CustomFactor' class as well as any of the datasets one wishes to use. Something like this.

# Imports needed to create and run pipeline  
from quantopian.pipeline import Pipeline, CustomFactor  
from quantopian.research import run_pipeline  

# Import any pipeline data we want to use. These are most common.  
from quantopian.pipeline.data.builtin import USEquityPricing  
from quantopian.pipeline.data import Fundamentals  # Morningstar  
from quantopian.pipeline.data import factset   # Factset

Next, create a subclass of CustomFactor as a new class. (see the docs https://www.quantopian.com/help#custom-factors)

class MyFactor(CustomFactor):  
    # assign any default input(s). not required but maybe convenient.  
    inputs = [Fundamentals.cash_and_cash_equivalents, USEquityPricing.close]  
    # assign a default window_length. again not required  
    window_length = 2  
    # factors can have multiple outs. if so, then give them names  
    outputs= ["cash_n_days_ago", "close_n_days_ago"]


    def compute(self, today, asset_ids, out, cash, close):  
        # do any logic here  
        # the inputs appear in the same order as assigned above  
        # inputs are numpy arrays.  
        # columns are the assets. rows are the days with the earliest first [0] and most recent last [-1]  
       out.cash_n_days_ago[:] = cash[0]  
       out.close_n_days_ago[:] = close[0]

The above custom factor will return the 'cash_and_cash_equivalents' and the 'close' price from 2 days ago. It needs to be instantiated when defining the pipeline like this

my_factor = MyFactor()  
cash_n_days_ago = my_factor.cash_n_days_ago  
close_n_days_ago = my_factor.close_n_days_ago

This shows how a custom factor can have multiple outputs. Typically one would only have a single output. This is just to show how.

Often one doesn't need to write a custom factor especially for the fundamental data. Typically one just wants the latest value without doing any manipulation. A factor can simply be built directly from the dataset column by using the latest method.

current_cash = Fundamentals.cash_and_cash_equivalents.latest

Finally, there are currently a couple of sources for fundamental data - Morningstar and Factset. Both can be used together if desired. Take a look at the documentation for available fields

https://www.quantopian.com/docs/data-reference/overview
https://www.quantopian.com/docs/data-reference/factset_fundamentals
https://www.quantopian.com/docs/data-reference/morningstar_fundamentals

Attached is a notebook which shows a custom factor in action. It's a modification of the one used in this post https://www.quantopian.com/posts/change-in-fundamental-data-in-research

Hope that helps.