I have a database of users posts in a social network, Using Pandas DataFrame I counted the monthly number of posts of each user, which resulted in a 2 column table for each user containing the month and number of posts. I want to calculate the monthly count correlation between different users, knowing that the monthly timelines are differents between each two users ( with some intersected months)
This is the code for creating a monthly timeline table (agg)
# Create an empty dataframe
df = pd.DataFrame()
# Create a column from the datetime variable
df['datetime'] = date_list
# Convert that column into a datetime datatype
df['datetime'] = pd.to_datetime(df['datetime'])
# Set the datetime column as the index
df['score'] = count
df.index = df['datetime']
# this is the table containing posts count for each month
agg = df['score'].resample('M').sum().to_frame()
So basically I have to apply a correlation function on two "agg" variables, but couldn't find an intuitive way to do it. Here is two examples of agg variable belonging to two different users:
First column : Month ,Second column : Number of posts.
User A
2018-04-30 39
2018-05-31 41
2018-06-30 19
2018-07-31 46
2018-08-31 61
2018-09-30 57
2018-10-31 33
2018-11-30 18
User B:
2017-11-30 0
2017-12-31 3
2018-01-31 0
2018-02-28 0
2018-03-31 22
2018-04-30 3
2018-05-31 11
here is a solution to calculus the pearson correlation:
import pandas as pd
data = """
datetime score
2018-04-30 39
2018-05-31 41
2018-06-30 19
2018-07-31 46
2018-08-31 61
2018-09-30 57
2018-10-31 33
2018-11-30 18
"""
datb = """
datetime score
2017-11-30 0
2017-12-31 3
2018-01-31 0
2018-02-28 0
2018-03-31 22
2018-04-30 3
2018-05-31 11
"""
dfa = pd.read_csv(pd.compat.StringIO(data), sep='\s+')
dfb = pd.read_csv(pd.compat.StringIO(datb), sep='\s+')
dfa['datetime'] = pd.to_datetime(dfa['datetime'])
dfb['datetime'] = pd.to_datetime(dfb['datetime'])
dfa.index = dfa['datetime']
dfb.index = dfb['datetime']
agga = dfa['score'].resample('M').sum().to_frame()
aggb = dfb['score'].resample('M').sum().to_frame()
print(agga,aggb)
#intersection of 2 dataframes on datetime
inter = agga.merge(aggb, on='datetime')
print(inter)
result = inter['score_x'].corr(inter['score_y'])
print(result)
dfa
score
datetime
2018-04-30 39
2018-05-31 41
2018-06-30 19
2018-07-31 46
2018-08-31 61
2018-09-30 57
2018-10-31 33
2018-11-30 18
dfb
score
datetime
2017-11-30 0
2017-12-31 3
2018-01-31 0
2018-02-28 0
2018-03-31 22
2018-04-30 3
2018-05-31 11
inter
score_x score_y
datetime
2018-04-30 39 3
2018-05-31 41 11
result
0.9999999999999999
If you want to use union:
union = pd.merge(agga, aggb, on='datetime', how='outer').fillna(0)
output of union:
score_x score_y
datetime
2018-04-30 39.0 3.0
2018-05-31 41.0 11.0
2018-06-30 19.0 0.0
2018-07-31 46.0 0.0
2018-08-31 61.0 0.0
2018-09-30 57.0 0.0
2018-10-31 33.0 0.0
2018-11-30 18.0 0.0
2017-11-30 0.0 0.0
2017-12-31 0.0 3.0
2018-01-31 0.0 0.0
2018-02-28 0.0 0.0
2018-03-31 0.0 22.0
a good link to understant merge