Delhivery Data Analysis

About Delhivery

Delhivery is the largest and fastest-growing fully integrated logistics provider in India as of Fiscal 2021. The company aims to build the operating system for commerce through a blend of world-class infrastructure, high-quality logistics operations, and cutting-edge engineering and technology capabilities.

The data team at Delhivery leverages vast datasets to enhance business intelligence, drive operational efficiency, and maintain profitability, creating a significant competitive edge.

---

Objective

The goal of this project is to process and analyze data generated by Delhivery's logistics operations to:

1. Clean, sanitize, and manipulate raw data to derive actionable insights.
2. Create useful features for the data science team to develop forecasting models.

---

Dataset

The dataset consists of records from Delhivery's logistics and operational data pipeline.

Key Features:

• data: Indicates if the record is training or testing data.
• trip_creation_time: Timestamp of trip creation.
• route_schedule_uuid: Unique identifier for a route schedule.
• route_type: Type of transportation (FTL, Carting).
  • FTL: Full Truck Load shipments, faster delivery as there are no intermediate pickups/drop-offs.
  • Carting: Delivery system using smaller vehicles (carts).
• trip_uuid: Unique identifier for a trip (a trip can involve multiple source and destination centers).
• source_center: ID of the trip's origin center.
• source_name: Name of the trip's origin center.
• destination_center: ID of the destination center.
• destination_name: Name of the destination center.
• od_start_time: Trip start time.
• od_end_time: Trip end time.
• start_scan_to_end_scan: Total time taken for delivery from source to destination.
• actual_distance_to_destination: Actual distance in kilometers between source and destination.
• actual_time: Cumulative time taken to complete the delivery.
• osrm_time: Time calculated by the Open-Source Routing Machine (OSRM) considering shortest paths and typical traffic conditions (cumulative).
• osrm_distance: Distance calculated by OSRM (cumulative).
• segment_actual_time: Time taken for a segment of the delivery.
• segment_osrm_time: OSRM-calculated time for a delivery segment.
• segment_osrm_distance: OSRM-calculated distance for a delivery segment.

Additional Fields:

Some fields with currently unclear meanings, like is_cutoff, cutoff_factor, cutoff_timestamp, and factor, are included for completeness and may be explored further.

---

Process Overview

1. Feature Engineering:

• Derived meaningful metrics such as:
  • time_diff_hours: Time difference between od_start_time and od_end_time.
  • Extracted components from timestamps (e.g., month, year, day of the week).
  • Split and standardized source and destination names into city, place code, and state.

2. Data Cleaning:

• Handled missing values using appropriate imputation techniques.
• Addressed outliers with boxplots and the IQR method.

3. Categorical Feature Handling:

• Applied one-hot encoding to variables like route_type for better interpretability in downstream models.

4. Normalization and Standardization:

• Used MinMaxScaler and StandardScaler for numerical columns to align features to a uniform scale.

---

Key Insights

1. Route Type Insights:

   • FTL routes are faster and more efficient for long distances compared to Carting.

2. Source and Destination Patterns:

   • High-frequency routes indicate key operational hubs that could benefit from resource optimization.

3. Time Efficiency:

   • Delivery times vary significantly by route type, season, and traffic conditions.

4. OSRM vs. Actual Metrics:

   • Discrepancies between OSRM-calculated and actual times/distances highlight areas for improving routing algorithms.

---

Tools and Libraries

This project utilized the following tools:

• Python:
  • Pandas for data manipulation.
  • Matplotlib and Seaborn for visualization.
  • Sklearn for preprocessing and scaling.
• Jupyter Notebook: For interactive analysis and documentation.

---

Repository Structure

• data/: Contains the dataset used for analysis.
• notebooks/: Jupyter Notebooks documenting the analysis process.
• visualizations/: Saved plots and charts.
• README.md: Overview of the project (this file).

---

Next Steps

Future directions for this project include:

1. Developing predictive models for delivery time and distance.
2. Investigating patterns in the unknown fields (is_cutoff, cutoff_factor, etc.).
3. Implementing clustering techniques to identify high-demand routes.

---

Acknowledgments

• Dataset Source: Provided by Scaler for this analysis.
• Python Libraries: Thanks to the open-source Python community for providing versatile data analysis tools.

---

License

This project is licensed for educational and non-commercial use only. If utilizing any part of this repository, please credit the author.