I have a pandas dataframe like following..
date item_id
2016-01-19 [188, 188]
2016-01-23 [188, 142]
2016-02-05 [188, 264]
2016-02-06 [273, 248, 191, 167, 238, 191]
2016-02-15 [320]
2016-02-17 [286]
2016-02-20 [164, 317]
In the above I want to calculate a ratio of No of times the item_id got repeated on different dates / no of unique item_id
So in above scenario item_id 188
repeated 3 times on 3 different days so the ratio will be 3/no of unique item_id
3/13
code to create a dataframe
buyer_id item_id date
261_23 188 2016-01-19
261_23 188 2016-01-19
261_23 188 2016-01-23
261_23 142 2016-01-23
261_23 188 2016-02-05
261_23 264 2016-02-05
261_23 273 2016-02-06
261_23 248 2016-02-06
261_23 191 2016-02-06
261_23 167 2016-02-06
261_23 238 2016-02-06
261_23 191 2016-02-06
261_23 320 2016-02-15
261_23 286 2016-02-17
261_23 164 2016-02-20
261_23 317 2016-02-20
df.groupby(['date','buyer_id'])['item_id'].apply(lambda x: x.tolist())
The set of the union of all unique items is
unique_items = set().union(*df.item_id.apply(set))
The number of appearances of each item is
num_appearances = [df.item_id.apply(lambda s: k in s).sum() for k in unique_items]
Therefore, the following will create a dictionary mapping each item to the ratio you asked:
dict((k, n / float(len(unique_items))) \
for (k, n) in zip(unique_items, num_appearances))
Example
import pandas as pd
df = pd.DataFrame({
'date': range(5),
'item_id': [[188, 188], [188, 142], [188, 264], [273, 248, 191, 167, 238, 191], [320]]})
unique_items = set().union(*df.item_id.apply(set))
>>> unique_items
{142, 167, 188, 191, 238, 248, 264, 273, 320}
num_appearances = [df.item_id.apply(lambda s: k in s).sum() for k in unique_items]
>>> num_appearances
[1, 1, 1, 1, 1, 1, 1, 3, 1]
>>> dict((k, n / float(len(unique_items))) \
for (k, n) in zip(unique_items, num_appearances))
{142: 0.1111111111111111,
167: 0.1111111111111111,
188: 0.33333333333333331,
191: 0.1111111111111111,
238: 0.1111111111111111,
248: 0.1111111111111111,
264: 0.1111111111111111,
273: 0.1111111111111111,
320: 0.1111111111111111}
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.