intervention-analysis

In [1]:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

In [2]:

df1 = pd.read_csv("anonymized_all.csv")
df2 = pd.read_csv("anonymized_fy24_only.csv")

In [3]:

df1.head()

	Unique Identifier	Start Date	Client Name	Job ID
0	537926	6/14/21	537926 Record	job-188418
1	476870	6/18/21	476870 Record	job-188751
2	458255	6/14/21	458255 Record	job-189026
3	479256	8/4/20	479256 Record	job-181914
4	472697	8/7/20	472697 Record	job-181916

In [4]:

df2.head()

	Unique Identifier	Start Date	Client Name	Job ID
0	581013	9/25/23	581013 Record	job-193451
1	583586	7/17/23	583586 Record	job-198299
2	585047	7/11/23	585047 Record	job-198331
3	576444	7/9/23	576444 Record	job-198365
4	585401	9/18/23	585401 Record	job-198436

In [5]:

def clean_df(df, date_col, drop_col):
    df[date_col] = pd.to_datetime(df[date_col])
    df = df.drop([drop_col], axis = 1)
    df.columns = df.columns.str.lower()
    
    return df

cleaned_df1 = clean_df(df1,'Start Date', 'Client Name')
cleaned_df2 = clean_df(df2,'Start Date', 'Client Name')

In [6]:

cleaned_df1.head()

	unique identifier	start date	job id
0	537926	2021-06-14	job-188418
1	476870	2021-06-18	job-188751
2	458255	2021-06-14	job-189026
3	479256	2020-08-04	job-181914
4	472697	2020-08-07	job-181916

In [7]:

def fiscal_year(date):
    if date.month >=7:
        return date.year + 1
    else:
        return date.year
    
cleaned_df1['fiscal year'] = cleaned_df1['start date'].apply(fiscal_year)
cleaned_df2['fiscal year'] = cleaned_df2['start date'].apply(fiscal_year)

In [8]:

cleaned_df1['fiscal year'].unique()

array([2021, 2022, 2023, 2024])

In [9]:

cleaned_df2['fiscal year'].unique()

array([2024])

In [10]:

# <!-- ```{python}
# merged_df = pd.merge(cleaned_all, cleaned_fy24, how = 'outer')
# merged_df
# ```



# ```{python}
# merged_df['fiscal year'] = merged_df['start date'].apply(get_fiscal_year)
# merged_df.head()
# ```
# ```{python}
# merged_df['fiscal year'].unique()
# ```

# ```{python}
# merged_df = merged_df[merged_df['fiscal year'].isin([2023, 2024])]
# merged_df
# ```

# ```{python}
# weekly_performance = (
#   merged_df
#   .groupby(['fiscal year', pd.Grouper(key='start date', freq='W')]).size()
#   .reset_index(name = 'weekly_performance')
# )

# weekly_performance
# ```

# ```{python}
# weekly_performance['cumulative_performance'] = (
#   weekly_performance
#   .groupby('fiscal year')['weekly_performance']
#   .cumsum()
# )

# weekly_performance
# ```
# ```{python}
# # Pivot the data
# weekly_performance['start date'] = weekly_performance['start date'].dt.strftime('%m-%d')

# pivoted_data = weekly_performance.pivot(index='start date', columns='fiscal year', values='cumulative_performance')
# pivoted_data.columns = [f'Cumulative Performance FY{year}' for year in pivoted_data.columns]

# pivoted_data.reset_index(inplace=True)

# pivoted_data
# ```
# ```{python}
# # Plotting
# fig, ax = plt.subplots(figsize=(10, 6))
# pivoted_data.plot(x='start date', ax=ax)

# ax.set_title('Cumulative Weekly Performance by Fiscal Year')
# ax.set_xlabel('Date')
# ax.set_ylabel('Cumulative Performance')
# ax.legend()

# plt.show()
# ```
# ```{python}
# weekly_performance.to_csv("weekly_performance.csv", index = False) 
# ``` -->

In [11]:

cleaned_previous = cleaned_df1[cleaned_df1['fiscal year'].isin([2024])]
cleaned_previous

	unique identifier	start date	job id	fiscal year
8381	581013	2023-09-25	job-193451	2024
8382	583586	2023-07-17	job-198299	2024
8383	585047	2023-07-11	job-198331	2024
8384	576444	2023-07-09	job-198365	2024
8385	585401	2023-09-18	job-198436	2024
...	...	...	...	...
11009	619903	2024-03-13	job-205994	2024
11010	602062	2024-03-26	job-206010	2024
11011	598480	2024-03-06	job-206013	2024
11012	615498	2024-03-03	job-206017	2024
11013	460343	2024-02-11	job-206027	2024

2633 rows × 4 columns

In [12]:

previous_weekly_performance = (
  cleaned_previous
  .groupby(['fiscal year', pd.Grouper(key='start date', freq='W')]).size()
  .reset_index(name = 'previous_weekly_performance')
)
previous_weekly_performance.head()

	fiscal year	start date	previous_weekly_performance
0	2024	2023-07-02	15
1	2024	2023-07-09	77
2	2024	2023-07-16	99
3	2024	2023-07-23	71
4	2024	2023-07-30	72

In [13]:

previous_weekly_performance['previous_cumulative_performance'] = (
  previous_weekly_performance
  .groupby('fiscal year')['previous_weekly_performance']
  .cumsum()
)
previous_weekly_performance.head()

	fiscal year	start date	previous_weekly_performance	previous_cumulative_performance
0	2024	2023-07-02	15	15
1	2024	2023-07-09	77	92
2	2024	2023-07-16	99	191
3	2024	2023-07-23	71	262
4	2024	2023-07-30	72	334

In [14]:

current_weekly_performance = (
  cleaned_df2
  .groupby(['fiscal year', pd.Grouper(key='start date', freq='W')]).size()
  .reset_index(name = 'current_weekly_performance')
)
current_weekly_performance.head()

	fiscal year	start date	current_weekly_performance
0	2024	2023-07-02	15
1	2024	2023-07-09	77
2	2024	2023-07-16	98
3	2024	2023-07-23	72
4	2024	2023-07-30	72

In [15]:

current_weekly_performance['current_cumulative_performance']= (
  current_weekly_performance
  .groupby('fiscal year')['current_weekly_performance']
  .cumsum()
)
current_weekly_performance.head()

	fiscal year	start date	current_weekly_performance	current_cumulative_performance
0	2024	2023-07-02	15	15
1	2024	2023-07-09	77	92
2	2024	2023-07-16	98	190
3	2024	2023-07-23	72	262
4	2024	2023-07-30	72	334

In [16]:

merged_df = pd.merge(previous_weekly_performance, current_weekly_performance, how = 'outer')
merged_df.head()

	fiscal year	start date	previous_weekly_performance	previous_cumulative_performance	current_weekly_performance	current_cumulative_performance
0	2024	2023-07-02	15	15	15	15
1	2024	2023-07-09	77	92	77	92
2	2024	2023-07-16	99	191	98	190
3	2024	2023-07-23	71	262	72	262
4	2024	2023-07-30	72	334	72	334

In [17]:

merged_df.to_csv("weekly_performance.csv", index = False)

# Plotting 1
# Set 'start date' as the index for better plotting
merged_df.set_index('start date', inplace=True)

fig, ax = plt.subplots(figsize=(12, 8))

# Width of the bars
bar_width = 0.35

# Positions of the bars
index = range(len(merged_df))

# Plotting 'previous_weekly_performance'
bars1 = ax.bar(index, merged_df['previous_weekly_performance'], width=bar_width, label='Previous Weekly Performance', color='b')

# Plotting 'current_weekly_performance' shifted right by bar_width to place next to the previous
bars2 = ax.bar([p + bar_width for p in index], merged_df['current_weekly_performance'], width=bar_width, label='Current Weekly Performance', color='r')

# Adding labels, title, and legend
ax.set_xlabel('Start Date')
ax.set_ylabel('Weekly Performance')
ax.set_title('Comparison of Previous and Current Weekly Performance')
ax.legend()

# Set x-ticks to be in the middle of the two bars for each date
ax.set_xticks([p + bar_width / 2 for p in index])
ax.set_xticklabels(merged_df.index.strftime('%Y-%m-%d'), rotation=45)

plt.show()


# Plotting 2
# Resetting the index to put 'start date' back as a regular column
merged_df.reset_index(inplace=True)


plt.figure(figsize=(10, 6))

# Plotting the 'previous_cumulative_performance'
plt.plot(merged_df['start date'], merged_df['previous_cumulative_performance'], label='Previous Cumulative Performance', marker='o')

# Plotting the 'current_cumulative_performance'
plt.plot(merged_df['start date'], merged_df['current_cumulative_performance'], label='Current Cumulative Performance', marker='o')
 
plt.title('Comparison of Previous and Current Cumulative Performance Over Time')
plt.xlabel('Start Date')
plt.ylabel('Cumulative Performance')
plt.legend()
plt.grid(True)
plt.xticks(rotation=45)  # Rotates the dates on the x-axis for better visibility
plt.tight_layout()  # Adjusts plot parameters to give some padding and prevent overlap

plt.show()

(a) The bar chart shows weekly performance trends for previous and current periods, highlighting fluctuations and potential seasonal patterns. This dense representation helps identify key variations but can be visually complex.

1-intervention-analysis.ipynb