Skip to main content

​
Overview

The application generates two types of metrics reports:
  1. Weekly Metrics - Aggregated trip statistics by week
  2. Monthly Metrics by Rate Code - Service volumes by month, rate code, and day type

​
Weekly Metrics

Weekly metrics provide time-series analysis of trip characteristics aggregated by ISO week.

​
Implementation

main.py:74-94
def generate_week_metrics(self):
    self.data['trip_time'] = self.data['tpep_dropoff_datetime'] - self.data['tpep_pickup_datetime']
    self.data['trip_time_in_seconds'] = self.data['trip_time'].dt.total_seconds()

    self.csv_df = self.data.groupby('year_week').agg(
        min_trip_time=('trip_time_in_seconds', 'min'),
        max_trip_time=('trip_time_in_seconds', 'max'),
        mean_trip_time=('trip_time_in_seconds', 'mean'),
        min_trip_distance=('trip_distance', 'min'),
        max_trip_distance=('trip_distance', 'max'),
        mean_trip_distance=('trip_distance', 'mean'),
        min_trip_amount=('total_amount', 'min'),
        max_trip_amount=('total_amount', 'max'),
        mean_trip_amount=('total_amount', 'mean'),
        total_services=('total_amount', 'count')
    ).reset_index()

    self.csv_df['percentage_variation'] = (
        self.csv_df['total_services'] - self.csv_df['total_services'].shift(1)
        ) / self.csv_df['total_services'].shift(1) * 100

​
Calculated Fields

Derived Column:
main.py:75-76
self.data['trip_time'] = self.data['tpep_dropoff_datetime'] - self.data['tpep_pickup_datetime']
self.data['trip_time_in_seconds'] = self.data['trip_time'].dt.total_seconds()
Aggregations:
  • min_trip_time - Shortest trip duration in the week (seconds)
  • max_trip_time - Longest trip duration in the week (seconds)
  • mean_trip_time - Average trip duration for the week (seconds)
Use Case: Analyze traffic patterns, identify rush hour impacts, detect service quality trends.
Source Column: trip_distance (miles)Aggregations:
  • min_trip_distance - Shortest trip distance in the week
  • max_trip_distance - Longest trip distance in the week
  • mean_trip_distance - Average trip distance for the week
Use Case: Track average trip lengths, identify long-haul vs. short-trip patterns, analyze service area coverage.
Source Column: total_amount (USD)Aggregations:
  • min_trip_amount - Lowest fare in the week
  • max_trip_amount - Highest fare in the week
  • mean_trip_amount - Average fare for the week
Use Case: Monitor revenue trends, detect pricing anomalies, analyze fare patterns.
Aggregation:
  • total_services - Total number of trips completed in the week
main.py:88
total_services=('total_amount', 'count')
Use Case: Track demand trends, measure service capacity utilization, identify growth patterns.
Uses count aggregation on total_amount column (any non-null column would work since data is already cleaned).

​
Percentage Variation

The week-over-week change in service volume:
main.py:91-93
self.csv_df['percentage_variation'] = (
    self.csv_df['total_services'] - self.csv_df['total_services'].shift(1)
    ) / self.csv_df['total_services'].shift(1) * 100
Formula: 
percentage_variation = ((current_week_services - previous_week_services) / previous_week_services) × 100
Example:
  • Week 1: 10,000 trips
  • Week 2: 10,500 trips
  • Percentage variation = ((10,500 - 10,000) / 10,000) × 100 = +5%
Interpretation
  • Positive values indicate growth in service volume
  • Negative values indicate decline in service volume
  • First week will have NaN (no previous week to compare)

​
Grouping Key: year_week

Trips are grouped by ISO week format:
main.py:69-70
self.data['year_dt'] = self.data['tpep_dropoff_datetime'].dt.year.astype(str)
self.data['week_dt'] = self.data['tpep_dropoff_datetime'].dt.isocalendar().week.astype(str).str.zfill(3)
self.data['year_week'] = self.data['year_dt'].str.cat(self.data['week_dt'], sep='-')
Format: YYYY-WWW (e.g., 2022-012 for week 12 of 2022)
Week numbers are zero-padded to 3 digits for consistent sorting.

​
Output Format

Exported to processed_data.csv with pipe delimiter:
main.py:135
self.csv_df.to_csv('processed_data.csv', sep='|', index=False)
Sample Output: 
year_week|min_trip_time|max_trip_time|mean_trip_time|min_trip_distance|max_trip_distance|mean_trip_distance|min_trip_amount|max_trip_amount|mean_trip_amount|total_services|percentage_variation
2022-001|60.00|5832.00|892.45|0.01|38.50|3.12|0.50|287.80|18.34|45623|
2022-002|60.00|6012.00|901.23|0.01|42.30|3.18|0.50|312.50|18.72|47891|4.97

​
Monthly Metrics by Rate Code

Monthly metrics segment trips by rate code type and day type (weekday vs. weekend).

​
Implementation

main.py:96-123
def generate_month_metrics(self):
    rate_code_id_dict = {
        'regular_df': 1,
        'jfk_df': 2,
        'other_df': -1
    }

    self.data['day_type'] = np.where(self.data['tpep_dropoff_datetime'].dt.dayofweek >= 5, 2, 1)
    self.data['RatecodeID'] = self.data['RatecodeID'].astype(int)

    for rc_id in rate_code_id_dict.keys():
        attr = getattr(self, rc_id)

        if rate_code_id_dict[rc_id] in [1, 2]:
            df = self.data[self.data['RatecodeID'] == rate_code_id_dict[rc_id]]
        else:
            df = self.data[(self.data['RatecodeID'] != 1) & (self.data['RatecodeID'] != 2)]

        df = df[['year_month', 'day_type', 'trip_distance', 'passenger_count']]

        df = df.groupby(['year_month', 'day_type']).agg(
            services=('trip_distance', 'count'),
            distances=('trip_distance', 'sum'),
            passengers=('passenger_count', 'sum')
        ).reset_index()

        attr = pd.concat([attr, df])
        setattr(self, rc_id, attr)

​
Day Type Classification

main.py:103
self.data['day_type'] = np.where(self.data['tpep_dropoff_datetime'].dt.dayofweek >= 5, 2, 1)
Day Type Values:
  • 1 = Weekday (Monday through Friday)
  • 2 = Weekend (Saturday and Sunday)
Logic:
  • dayofweek >= 5 means Saturday (5) or Sunday (6)
  • dayofweek < 5 means Monday (0) through Friday (4)
Why Segment by Day Type?Weekday and weekend trips have different characteristics:
  • Weekdays: More commuter/business trips, rush hour patterns
  • Weekends: More leisure/tourist trips, different time distributions

​
Rate Code Segmentation

Three separate DataFrames are created:
main.py:109-110
if rate_code_id_dict[rc_id] in [1, 2]:
    df = self.data[self.data['RatecodeID'] == rate_code_id_dict[rc_id]]
Stored in: self.regular_dfStandard metered trips - the majority of yellow taxi services.
main.py:109-110
if rate_code_id_dict[rc_id] in [1, 2]:
    df = self.data[self.data['RatecodeID'] == rate_code_id_dict[rc_id]]
Stored in: self.jfk_dfJFK airport flat-rate trips - analyzed separately due to unique pricing and distance patterns.
main.py:111-112
else:
    df = self.data[(self.data['RatecodeID'] != 1) & (self.data['RatecodeID'] != 2)]
Stored in: self.other_dfAll other rate codes: negotiated fares, Nassau/Westchester, group rides, etc.

​
Aggregated Metrics

For each rate code segment, trips are grouped by [year_month, day_type]:
main.py:116-120
df = df.groupby(['year_month', 'day_type']).agg(
    services=('trip_distance', 'count'),
    distances=('trip_distance', 'sum'),
    passengers=('passenger_count', 'sum')
).reset_index()
Metrics:
​
services
int
Total number of trips for the month/day_type combination.Uses count on trip_distance (any non-null column works).
​
distances
float
Total miles traveled for all trips in the month/day_type combination.Sum of all trip_distance values.
​
passengers
int
Total passengers served for all trips in the month/day_type combination.Sum of all passenger_count values.

​
Grouping Keys

year_month:
main.py:67
self.data['year_month'] = self.data['tpep_dropoff_datetime'].dt.strftime('%Y-%m')
Format: YYYY-MM (e.g., 2022-01) day_type:
  • 1 = Weekday
  • 2 = Weekend

​
Output Format

Exported to Excel with separate sheets:
main.py:138-143
common_columns = ['year_month', 'day_type', 'services', 'distances', 'passengers']
with pd.ExcelWriter("processed_data.xlsx", engine="openpyxl") as writer:
    self.jfk_df[common_columns].to_excel(writer, sheet_name="JFK", index=False)
    self.regular_df[common_columns].to_excel(writer, sheet_name="Regular", index=False)
    self.other_df[common_columns].to_excel(writer, sheet_name="Others", index=False)
File: processed_data.xlsx Sheets:
  • JFK - JFK airport trips
  • Regular - Standard rate trips
  • Others - All other rate codes
Sample Output (Regular sheet): 
year_month | day_type | services | distances | passengers
2022-01    | 1        | 234567   | 731234.56 | 289012
2022-01    | 2        | 98234    | 312456.78 | 121345
2022-02    | 1        | 241890   | 755432.10 | 298765
2022-02    | 2        | 101234   | 321098.45 | 125678

​
Data Formatting

Before export, metrics are formatted:
main.py:126-131
def format_data(self):
    self.csv_df = self.csv_df.round(2)

    self.jfk_df = self.jfk_df.reset_index()
    self.regular_df = self.regular_df.reset_index()
    self.other_df = self.other_df.reset_index()
  • Weekly metrics rounded to 2 decimal places
  • Monthly metrics indexes reset for clean export

​
Analysis Use Cases

​
Weekly Metrics

  • Trend analysis over time
  • Week-over-week growth tracking
  • Seasonal pattern detection
  • Service quality monitoring (avg duration/distance/fare)

​
Monthly Metrics

  • Rate code performance comparison
  • Weekday vs. weekend demand analysis
  • Passenger volume trends
  • Distance-based service characterization
  • Revenue segment analysis
Combining Both MetricsWeekly metrics provide granular time-series data, while monthly metrics offer segmented business intelligence. Together, they enable comprehensive operational and strategic analysis.

​
Next Steps

Build docs developers (and LLMs) love

Get started for freeTalk to us