Uber Data Project

📊 Click here for data set information

About Dataset

Uber Ride Analytics Dataset 2024

This comprehensive dataset contains detailed ride-sharing data from Uber operations for the year 2024, providing rich insights into booking patterns, vehicle performance, revenue streams, cancellation behaviors, and customer satisfaction metrics.

Dataset Overview

The dataset captures 148,770 total bookings across multiple vehicle types and provides a complete view of ride-sharing operations including successful rides, cancellations, customer behaviors, and financial metrics.

Key Statistics:

• Total Bookings: 148.77K rides

• Success Rate: 65.96% (93K completed rides)

• Cancellation Rate: 25% (37.43K cancelled bookings)

• Customer Cancellations: 19.15% (27K rides)

• Driver Cancellations: 7.45% (10.5K rides)

Data Schema

The dataset contains the following columns:

Column Name	Description
Date	Date of the booking
Time	Time of the booking
Booking ID	Unique identifier for each ride booking
Booking Status	Status of booking (Completed, Cancelled by Customer, Cancelled by Driver, etc.)
Customer ID	Unique identifier for customers
Vehicle Type	Type of vehicle (Go Mini, Go Sedan, Auto, eBike/Bike, UberXL, Premier Sedan)
Pickup Location	Starting location of the ride
Drop Location	Destination location of the ride
AVG VTAT	Average time for driver to reach pickup location (in minutes)
AVG CTAT	Average trip duration from pickup to destination (in minutes)
Cancelled Rides by Customer	Customer-initiated cancellation flag
Reason for cancelling by customer	Reason for customer cancellation
Cancelled Rides by Driver	Driver-initiated cancellation flag
Driver Cancellation Reason	Reason for driver cancellation
Incomplete Rides	Incomplete ride flag
Incomplete Rides Reason	Reason for incomplete rides
Booking Value	Total fare amount for the ride
Ride Distance	Distance covered during the ride (in km)
Driver Ratings	Rating given to driver (1-5 scale)
Customer Rating	Rating given by customer Method used for payment (UPI, Cash, Credit Card, Uber Wallet, Debit Card)

Vehicle Fleet Coverage

Vehicle Type	Total Bookings	Success Rate	Avg Distance	Total Distance
Auto	12.88M	91.1%	25.99 km	602K km
eBike/Bike	11.46M	91.1%	26.11 km	537K km
Go Mini	10.34M	91.0%	25.99 km	482K km
Go Sedan	9.37M	91.1%	25.98 km	433K km
Premier Sedan	6.28M	91.2%	25.95 km	292K km
UberXL	1.53M	92.2%	25.72 km	72K km

Revenue Distribution by Payment Method

• UPI: Highest contributor (~40% of total revenue)
• Cash: Second highest (~25% of total revenue)
• Credit Card: ~15% of total revenue
• Uber Wallet: ~12% of total revenue
• Debit Card: ~8% of total revenue

Cancellation Patterns

Customer Cancellation Reasons:

• Wrong Address: 22.5%
• Driver Issues: 22.4%
• Driver Not Moving: 22.2%
• Change of Plans: 21.9%
• App Issues: 11.0%
• Driver Cancellation Reasons:
• Capacity Issues: 25.0%
• Customer Related Issues: 25.3%
• Personal & Car Issues: 24.9%
• Customer Behavior: 24.8%

Rating Analysis

• Customer Ratings: Consistently high across all vehicle types (4.40-4.41)
• Driver Ratings: Slightly lower but stable (4.23-4.24)
• Highest Rated: Go Sedan (4.41 customer rating)
• Most Satisfied Drivers: UberXL category (4.24 rating)

Data Quality

• Completeness: Comprehensive coverage with minimal missing values
• Consistency: Standardized vehicle types and status categories
• Temporal Coverage: Full year 2024 data with daily granularity
• Geographic Scope: Multiple pickup and drop locations
• Balanced Distribution: Good representation across all vehicle types and time periods.

1. Data Cleaning Process

For the Uber Ride Bookings dataset, I designed and implemented a reproducible data cleaning pipeline using Python’s pandas and NumPy libraries. The process ensured the dataset was reliable, consistent, and ready for analysis.

Data Loading & Validation

• Imported the dataset (original_uber_ride_bookings.csv) with robust error handling to ensure smooth execution even if the file was missing or misnamed.

• Conducted initial exploratory checks (head(), info()) to inspect data structure, column types, and missing values.

Handling Missing Values

• Replaced null entries in numeric columns with the mean of each column. This imputation method preserved the overall distribution and ensured key metrics like averages remained representative.

Duplicate Removal

• Removed duplicate rows to prevent skewed analyses and maintain dataset integrity.

Basic Statistical Validation

• Generated descriptive statistics (describe()) to confirm distributions and validate that cleaning steps had worked.

• Performed a targeted check on the Booking_Value column by computing its mean as a sanity check of numerical accuracy.

Data Export & File Management

• Exported the cleaned dataset as cleaned_uber_data.csv for use in subsequent analysis and visualization tasks.

• Renamed the original dataset file to original_uber_data.csv to avoid confusion and maintain clear version control.

✨Result: A structured, fully cleaned dataset with missing values imputed, duplicates removed, and key metrics validated to ensure accurate, reliable insights for downstream analysis and reporting.

2. Correlation Analysis

• Calculated the Pearson correlation between Driver_Ratings and Booking_Value.
  • Result: -0.0002 → essentially no linear relationship.
• Visualized via a scatter plot, confirming no clear pattern.
• Created a full correlation matrix for all numeric variables, finding no strong linear relationships.
  ✨Result: linear regression was not appropriate for this dataset.

3. Logistic Regression Testing

• Reframed the problem: Does driving a Sedan predict high-value rides?
  • Defined high_value as booking values above the dataset average ($508.29).
  • Encoded Vehicle_Type as a binary variable (is_sedan).
• Built features: X = [is_sedan, Booking_Value] and target y = high_value.
• Split the data into training and test sets using stratified sampling to preserve class balance.

4. Model Training & Evaluation

• Trained a Logistic Regression model using scikit-learn.
• Evaluated performance with:
  • Confusion matrix
  • Classification report (precision, recall, F1)
  • ROC Curve to visualize model discrimination ability.
• Extracted coefficients to interpret feature importance (e.g., whether Sedans increase odds of high-value rides).

5. Findings

• Correlation analysis showed no strong linear relationships across numeric features.
• Logistic regression revealed whether is_sedan carried predictive power for high-value rides.
• Additional checks of class balance (value_counts, group means) ensured the insights were valid and not artifacts of imbalance.

---

✨ This workflow demonstrates a full EDA → feature engineering → modeling → interpretation pipeline using pandas, NumPy, Matplotlib, and scikit-learn.