Machine Learning to Predict Uber Surge Price

Uber, the ride-sharing tech giant, aims to provide the best customer experience possible. When using a ride-sharing service, the price is never fixed. For the same starting point and destination, customers observe that the fare always varies. Uber's prices are affected by the supply and demand of rides at a given time. The price of a ride can depend on many factors from the weather to an extensive amount of people summoning a ride at the same time. When the demand for rides increases, the price for the ride increaes to ensure that people who need a ride can get one- according to Uber. The intention behind this is that the people who need rides "more" than others will pay the extra surge price whereas others who are in less of a rush can wait for a lower price. Even though the datasets below also contain data for Lyft, we will focus on Uber. This is currently a problem because people will end up paying way more on an Uber ride than they really should be paying.

Citations

References Delaney, D. (2016, May 29). Surge pricing: Why a rainy day ride will cost you more. Tennessean.com. https://www.tennessean.com/story/money/2016/05/29/surge-pricing-why-rainy-day-ride-cost-you-more/84922758/

Helling, B. (2023). Surge Pricing: What It Is & How It Works For Riders & Drivers. Ridester.com. https://www.ridester.com/surge-pricing/

How Surge Pricing Works. (n.d.). Uber.com. Retrieved February 27, 2023, from https://www.uber.com/us/en/drive/driver-app/how-surge-works/

Datasets from Kaggle

import pandas as pd

rides_df = pd.read_csv('cab_rides.csv')

rides_df.head()

	distance	cab_type	time_stamp	destination	source	price	surge_multiplier	id	product_id	name
0	0.44	Lyft	1544952607890	North Station	Haymarket Square	5.0	1.0	424553bb-7174-41ea-aeb4-fe06d4f4b9d7	lyft_line	Shared
1	0.44	Lyft	1543284023677	North Station	Haymarket Square	11.0	1.0	4bd23055-6827-41c6-b23b-3c491f24e74d	lyft_premier	Lux
2	0.44	Lyft	1543366822198	North Station	Haymarket Square	7.0	1.0	981a3613-77af-4620-a42a-0c0866077d1e	lyft	Lyft
3	0.44	Lyft	1543553582749	North Station	Haymarket Square	26.0	1.0	c2d88af2-d278-4bfd-a8d0-29ca77cc5512	lyft_luxsuv	Lux Black XL
4	0.44	Lyft	1543463360223	North Station	Haymarket Square	9.0	1.0	e0126e1f-8ca9-4f2e-82b3-50505a09db9a	lyft_plus	Lyft XL

Data Dictionary:

• distance: distance between source and destination
• cab_type: Uber or Lyft
• time_stamp: epoch time when data was queried
• destination: destination of the ride
• source: the starting point of the ride
• price: price estimate for the ride in USD
• surge_multiplier: the multiplier by which price was increased, default 1
• id: unique identifier
• product_id: uber/lyft identifier for cab-type
• name: visible type of the cab: eg: Uber Pool, UberXL

weather_df = pd.read_csv('weather.csv')

weather_df.head()

	temp	location	clouds	pressure	rain	time_stamp	humidity	wind
0	42.42	Back Bay	1.0	1012.14	0.1228	1545003901	0.77	11.25
1	42.43	Beacon Hill	1.0	1012.15	0.1846	1545003901	0.76	11.32
2	42.50	Boston University	1.0	1012.15	0.1089	1545003901	0.76	11.07
3	42.11	Fenway	1.0	1012.13	0.0969	1545003901	0.77	11.09
4	43.13	Financial District	1.0	1012.14	0.1786	1545003901	0.75	11.49

Data Dictionary

• temp: temperature in F
• location: location name
• clouds: clouds
• pressure: pressure in mb
• rain: rain in inches for the last hr
• time_stamp: epoch time when row data was collected
• humidity: humidity in %
• wind: wind speed in mph

These two sets of data will be used to solve the problem by building a machine learning-based model that predicts the serge multipler- specifically based on the various weather conditions. The machine-learning based model that seems to be the best fit is ordinal regression/classification. This project will allow users to look at an area and see if there's currently a surge. This will be beneficial to users as they can choose to avoid this price by either waiting to call the Uber until the surge goes down or walking a bit to a starting position that is not in a surge zone.