capstone_project_starbucks_5.py

# Reference
offer_reference = {}
offer_duration_reference = {}

for i,j in zip(portfolio['id'], portfolio['offer_type']):
    offer_reference[i] = j

for i,j in zip(portfolio['id'], portfolio['duration']):
    offer_duration_reference[i] = j*24
    
# Set the list to store all the data
full_data = []

# Iterate through each person
for person in tqdm(list(transcript['person'].unique())):
    
    not_completed = {}
    received = []
    active = []
    total_data = {}
    information = []

    # Iterate through each person activity
    for index, row in transcript[transcript['person'] == person].iterrows():

        if row['event'] == 'offer received':

            # Everytime there is an offer received, do this
            received.append(row['offer_id'])
            key = row['offer_id'] + '-' + str(received.count(row['offer_id']))
            not_completed[key] = row['time']
            total_data[key] = [row['person'], row['offer_id'], 0, 0, 0, 0] 


        if row['event'] == 'offer viewed':

            # If the customers have seen the informational offer
            if offer_reference[row['offer_id']] == 'informational':
                information.append(row['offer_id'])

            # Everytime the offer is viewed, do this
            active = list(filter(lambda x: x.split('-')[0] == row['offer_id'], list(not_completed.keys())))

            # If there is only one offer_id active
            if len(active) == 1:

                # Only change the value if the offer is not completed yet
                if active[0] in not_completed:                    
                    total_data[active[0]][2] = 1

            # If there are more than one offer_id active
            else:
                for offer_id in active:
                    if (row['time'] - not_completed[offer_id]) < offer_duration_reference[row['offer_id']]:
                        if total_data[offer_id][2] == 1:
                            continue
                        total_data[offer_id][2] = 1
                        break


        if row['event'] == 'offer completed':

            # If the users completed the offer and have seen the informational offer
            info = False
            if len(information) > 0:
                info = True

            # Everytime the offer is completed, do this
            active = list(filter(lambda x: x.split('-')[0] == row['offer_id'], list(not_completed.keys())))

            # If there is only one offer_id active
            if len(active) == 1:
                total_data[active[0]][3] = 1
                total_data[active[0]][5] = row['time'] - not_completed[active[0]]
                not_completed.pop(active[0])

                if info:
                    total_data[active[0]][4] = 1

                continue

            # If there is more that one offer_id active
            else:
                for offer_id in active:
                    if (row['time'] - not_completed[offer_id]) < offer_duration_reference[row['offer_id']]:
                        total_data[offer_id][3] = 1
                        total_data[offer_id][5] = row['time'] - not_completed[offer_id]
                        not_completed.pop(offer_id)

                        if info:
                            total_data[offer_id][4] = 1

                        break

    for index, value in total_data.items():
        full_data += [value]

        
# Create a dataframe based on the compile result
compiled_data = pd.DataFrame(full_data, columns=['person', 'offer_id', 'viewed', 'completed', 'view_information', 'time_completed'])

# Merge with the portfolio and profile dataframe
compiled_data_merged = compiled_data.merge(portfolio, left_on='offer_id', right_on='id').drop(columns=['id'])
complete_data = compiled_data_merged.merge(profile, left_on='person', right_on='id').drop(columns=['id'])