inDrive Lahore — Trip Analysis

Real-World Data Science Project | BS Artificial Intelligence | Superior University

---

Problem Statement

inDrive riders in Lahore suffer from fare blindness — no data to judge whether the app's suggested fare is fair, what time gives cheapest rides, or which vehicle gets the fastest driver response. This project analyzes 498 real trips + 42 user survey responses to surface actionable insights.

3 Problems Solved

Problem	Finding
Fare Blindness	Riders overpay by 12.6% above app suggestion consistently
Time Trap	Night fares are 60% higher than morning fares
Ride Type Confusion	Cars get 3.94 avg bids vs rickshaws at 2.48

---

Project Structure

inDrive_Lahore_Analysis/
├── data/
│   ├── trips.csv           # 498 manual trip logs
│   └── survey.csv          # 42 user survey responses
├── data_cleaning.py        # Step 1: Clean & preprocess both datasets
├── eda_analysis.py         # Step 2: Generate 7 EDA plots
├── model.py                # Step 3: Fare prediction model
├── app.py                  # Streamlit portfolio dashboard
├── requirements.txt
└── README.md

---

How to Run

1. Install dependencies

pip install -r requirements.txt

2. Add your data files

Place your CSVs in the data/ folder:

• data/trips.csv (main trip log)
• data/survey.csv (user survey)

3. Run data cleaning

python data_cleaning.py

4. Run EDA (generates plots)

python eda_analysis.py

5. Run model

python model.py

6. Launch Streamlit app (main deliverable)

streamlit run app.py

---

Data Description

trips.csv (498 rows)

Column	Description
Date	Trip date
Team Member Name	Data collector
Time Category	Time slot (Morning/Evening/Night etc.)
Pickup Area	Origin zone in Lahore
Dropoff Area	Destination zone
Ride Type	Car / Bike / Rickshaw
Suggested Fare (PKR)	inDrive app suggestion
Team Member Offer (PKR)	Fare offered by rider
Final Accepted Fare (PKR)	Actual fare paid
Num of Driver Bids	Driver responses received
Weather Condition	Clear / Cloudy / Rain
Traffic Level	Low / Moderate

survey.csv (42 rows)

User perception survey covering: demographics, usage frequency, fare perception, peak hour awareness, frustrations, and comparison with Careem/Uber.

---

Key Findings

1. Negotiation Gap: Final fares consistently exceed suggested fares by 11–15% across all ride types
2. Night Dominance: 38% of trips occur at night (6–10 PM) but fares are 60% higher
3. Supply Imbalance: Cars receive most driver bids (3.94 avg); rickshaws least (2.48 avg)
4. User Frustration: Long wait & driver rejections are tied as top complaints (survey)
5. Competitive Edge: 90% of users find inDrive same or cheaper than Careem/Uber

Model Performance

• Algorithm: Random Forest Regressor
• Features: Suggested Fare, Driver Bids, Ride Type, Time, Weather, Traffic
• R² Score: ~0.97 (97% variance in final fare explained)
• Key Insight: Suggested Fare is the strongest predictor; time of day is second

---

Developed by Syed Tahir · BS Artificial Intelligence (4th Semester) · Superior University Lahore