This document presents a study on reducing surge pricing for cab hailing services during rush hours. It analyzes data from 58,000 taxi trips over 28 days to identify rush hour periods between 6-10pm Monday through Friday with high demand. An open network Poisson model is used to simulate driver and rider arrival rates. The model shows communicating rush hour timing and locations to drivers can decrease surge pricing by up to 10% by motivating more drivers to continue working during these periods. However, the assumptions require validation and driver psychology was not fully considered. Future work could focus on demand-supply equilibrium and behavioral factors.
How to Transform Clinical Trial Management with Advanced Data Analytics
Surge pricing mc_presentation
1. REDUCING SURGE PRICES FOR
CAB HAILING SERVICE
Presented by
Muktanidhi Dhotrad, Rohit Khurana, Sandeep Girada
Under the Guidance of
Dr. Yunan Liu
Yining Huang
Apr 2019
Financial Mathematics
2. Problem
Statement
Financial Mathematics
Problem Statement:
• How to bring down the
surge pricing during
rush hours to an
acceptable level?
Hypothesis:.
• By communicating to
the drivers about the
Rush hours timing and
geographical location,
surge pricing can be
brought down
• Here communication is
passive, i.e. at the start
of the day/week and
not during the rush
hours.
Available
Data
28 Days/4
Weeks
58k trips
Boundary
Conditions
Rush Hours
(6 pm -10
pm)
Days: M-F
MC
Techniques
Used
Open
Model
NHPP
Scope
3. Data Exploration, Cleaning
& Aggregation
• Classification of 58k trips into 168
Categories (24 hours * 7days) -
introduction of day_hour_factor
• Definition & Calculation of
– Rush Hours
– Arrival Rates of the Customers
(Requests/ Demand)
– Arrival Rate of the Drivers
(Supply_hours)
• Weekday rush hour comprises
– 8:00 am
– Hours between 5:00 pm and 10:00 pm
• Weekend rush hour comprises
– Hours between 4:00 pm and 10:00 pm
and 12:00 am
4. Key Patterns
0
1
2
3
4
5
Rider Cancelled Trips Rest of the Trips
Average ETA
73%
27%
Surge Multiplier
(=1) >1
AA BB CC DD
CC 15.14% 6.12% 73.01% 5.74%
𝜆𝑑
𝜆𝑟
• During Rush Hours (6 pm-10 pm, M-F)
• 90% of drivers leaves after the first trip
• 99.94% of the time Trips have been fulfilled
(matched)
• ~10% trips are cancelled
Rates/ per hour
(pm)
6 to 7 7 to 8 8 to 9 9 to 10
Supply rate 2.75 2.57 2.12 2.29
Demand rate 3.2 3 2.75 2.32
5. Definitions
Financial Mathematics
• Supply _hours: Total hours all partners (drivers) were online, en route, or on trip in the given hour
• Demand/ Requests: Total requests in the given hour
• Surge Pricing: When Supply > Demand, leads to surge pricing
• List of Service Times – Ts
• Total number of Drivers active in the System – N
• Model
• Open Network (p=0.9)
• Probability of 90 % drivers leaving the system after completion of 1 trip in a day
• With (1-p=0.1) probability of 10% returning back to system
6. Project Flow Diagram
Supply Function
• Poisson Distribution
• Arrival time of next
driver – td
Demand Function
• Poisson Distribution
• Arrival time of next
rider – tr
𝜆𝑑
𝜆r
Matching Function
• Recording of event
• Matching of
Requests and
Availability
Surge Decision
• Modelled using
Logistic
Regression
• Output as 1 or 0
indicating the
surge
Output Register
• Recording
1
2
3
4
7. Financial Mathematics
Open Network Model - Flow Chart
t= 0, n=5, T=240
Ts={list}, td, tr
Cap
Initialization
Min (Ts, td, tr)
t = tr
Generate ts
Ts.append(ts) Cancel Request
t < T
n + len(Ts) < Cap
n = n+1, t = td
Generate td
t = min(Ts{list})
Generate U
U < p
n = n+1
Delete(min(Ts{list}))
Tstrtd
n >0
Yes No
9. Financial Mathematics
Results
Fac=1
Cap p = 0.9 p = 0.8 p = 0.75 p = 0.7
15 0.63 0.6 0.587 0.566
16 0.61 0.573 0.552 0.523
17 0.588 0.535 0.449 0.46
18 0.566 0.5 0.46 0.423
• Target: Bring down surge to 50%
• Simulation window 6:00 pm – 10:00 pm on
Monday
# of
Drivers
Drivers motivated to continue in
the system
Fac=1.1
Cap p = 0.9 p = 0.8 p = 0.75 p = 0.7
15 0.61 0.578 0.545 0.52
16 0.57 0.54 0.48 0.44
17 0.56 0.51 0.45 0.41
18 0.53 0.49 0.46 0.374
Drivers motivated to continue in
the system
12. • Model uses lot of assumption, that needs to be validated.
• Psychological behavior of the partner/ drive does play a role and that
requires attention.
• Possibility of breach of Demand & Supply Equilibrium to be studied.
Future Enhancements