On Sat, 1 Oct 2016 18:12:29 -0700 (PDT), 38016226...@gmail.com wrote:
> I am trying to print a simple decision tree for my homework.
> The answer must keep in this format:
>
> Top 7,4,0.95
> career gain = 100
> 1.Management 2, 3, 0.9709505944546686
> 2.Service 5, 1, 0.6500224216483541
> location gain = 100
> 1.Oregon 4, 1, 0.7219280948873623
> 2.California 3, 3, 1.0
> edu_level gain = 100
> 1.High School 5, 1, 0.6500224216483541
> 2.College 2, 3, 0.9709505944546686
> years_exp gain = 100
> 1.Less than 3 3, 1, 0.8112781244591328
> 2.3 to 10 2, 1, 0.9182958340544896
> 3.More than 10 2, 2, 1.0
>
> Here is my code:
> features={'edu_level':['High School',
'College'],
'career': ['Management',
'Service'],
'years_exp':['Less than 3',
'3 to 10',
'More than 10'],
'location':['Oregon',
'California']}
>
> print('Top 7,4,0.95')
> for key in features:
> print('{} gain = {}'.format(key,100))
> attributes_list=features[key]
> kargs={}
> for i in range(len(attributes_list)):
> kargs[key]=attributes_list[i]
> low=table.count('Low',**kargs)
> high=table.count('High',**kargs)
> print('\t{}.{} {}, {}, {}'.format(
i+1,attributes_list[i],low,high,entropy(low,high)))
>
> I set all the gain as 100 now.But actually the gain must calculate
> with the data below. For example, the career gain need the data of
> 'Management' and 'Service'. I don't know how to do. or Anyone can
> provide me a better logic?
I interpret your question as meaning that the value that you
print after "gain =" should depend on features[key]. To do that,
you'll need to insert a line resembling
gain = gain_from_features(features[key])
before the print statement. You'll have to write the gain_from_features
function, and provide it with the numbers from which it will
compute the gain.
As a stylistic suggestion, note that Python allows you to break
your "features=" line into a more readable format, as I have done
above.
Another stylistic suggestions:
for key, attributes_list in features.iteritems():
--
To email me, substitute nowhere->runbox, invalid->com.
--
https://mail.python.org/mailman/listinfo/python-list
