[Analysis] DataFrame Code

Sorting

# Sort by multiple cols
df = df.sort_values(by=['Col_A', 'Col_B']).reset_index(drop=True)

# Sort descending
df = df.sort_values(by='Col_A', ascending = False)

nunique() vs unique() vs value_counts()

df['Col_A'].nunique()

• It calculates the number of unique Col_A

df['Col_A'].unique()

• It shows all unique Col_A rows

df['Col_A'].value_counts()

• It returns a series of containing counts of unique values.

Drop column

df = df.drop(columns=['Col_A,', 'Col_B'])

Min & Max of the date

df['Col_date'] = pd.to_datetime(df['Col_date'])

min_date = df['Col_date'].min()
max_date = df['Col_date'].max()

duration = max_date - min_date

Detail	Date
Min_Date	2022-01-01
Start_Date	2023-12-31
Duration	729 days

.groupby()

df['time'] = pd.to_datetime(recent['time'])
df['Year'] = df['time'].dt.year
df['Quarter'] = df['time'].dt.to_period('Q') # this makes 2022Q1

quarterly_sub = df.groupby(['Year', 'Quarter'])['user_id'].nunique().reset_index()
quarterly_sub['Quarter'] = quarterly_sub['Quarter'].astype(str)
quarterly_sub.columns = ['Year', 'Quarter', '# of sub_user']

print(quarterly_sub)

Year	Quarter	# of sub_user
2022	2022Q1	17628
2022	2022Q2	15280
2022	2022Q3	10722
2022	2022Q4	14757
2023	2023Q1	24983
2023	2023Q2	17241
2023	2023Q3	21180
2023	2023Q4	23342

• It groups the df by the Year and Quarter cols, and then it counts the # of unique user_id values in each group. This calculation determines how many distinct users perfromed the action logged in 'time' within each quarter.
• It converts it into a string for easier readability and use in displays or reports.

.astype()

df.astype(dtype, copy=True, errors='raise')

drop_duplicates() vs .duplicated()

df.drop_duplicates(subset=None, *, keep='first', inplace=False, ignore_index=False)

# remove duplicates on specific col, use subset
df.drop_duplicates(subset=['Col_A'])
# remove duplicates and keep last

• It returns dataframe with duplicate rows removed.

df.duplicated(subset=None, keep='first')

# By default, for each set of duplicated values, the first occurrence is set on False and all others on True. 
df.dulicated()