Topic: CO2 emissions and electric cars Due 10 hours, Report and Presentation I attached a similar type of report Thank you!Samp

DS 408: Project Proposal Report Simulating Ridesharing Apps (Supply & Demand)

Executive Summary

Ridesharing apps have grown exponentially in the last 15 years, even dominating over traditional taxi markets. One major issue with these apps that can create detractors is long waiting times. This is a problem we personally experienced when using these platforms. Waiting times can lead to a drop in revenue, consumer satisfaction and user retention. Since there are multiple causes behind them and we don’t have access to their backend, we’ve decided to simulate them and see if we can make improvements or resolve bottlenecks.

We are setting up a simulation model that matches supply & demand (i.e. drivers & riders) based on an approximate location, combined with open-source traffic data from a ridesharing giant. Our goal is to identify the bottleneck and implement a continuous solution that will create a better user experience for both the rider and the driver. We can use Excel to collect, organize and process data and SIGMA to design a model with varying parameters.

If successful, we can attempt to generalize and extrapolate our supply & demand matching algorithm to other use cases, and possibly sharing it in the open-source community.

How Ridesharing Operates

When a rider submits a request, it’s submitted in the back-end and a matchmaking process begins between all currently active drivers within a reasonable distance. All active drivers will receive a notification and have the option to accept or deny (or ignore) the request. To make things more complicated, riders can choose between different driver options; carpooling, single car, luxury cars, large vans, EV’s or even cars with disability provisions. So these types of drivers, along with who’s currently on the road, will create a supply distribution that we’ll need to simulate in SIGMA. Once en route, it may be possible for the driver to accept another rider, and may make one or more stops before reaching the final destination.

Our Simulation Approach

We will start with the basic model of accepting a request from a driver and matching it with the current supply of drivers. Once we set this up, we can add more factors, such as different driver options, roadblocks or traffic jams, surge pricing and disruptions caused by events.

We will assess where drivers are affected by high demand, and where consumers run the problem of not having a driver. The main objective is to identify the factors which cause extended wait times.

Parameters:

Distance Availability Type of vehicle Traffic jams Surge pricing

1. Using Excel

We will use the Uber Movement portal to obtain open-source traffic data in CSV format, which can be imported and organized into Excel. We can also use Excel to create random distributions for the riders’ demand and drivers’ supply. This will help us determine best and worst case scenarios, as well as peak times and bottlenecks which cause long wait times.

2. Using Sigma

We will use SIGMA to create queues of riders and drivers, and matching them based on the driver’s distance, time of day, and availability. We will run simulations for varying demographic areas, from dense metropolitan to rural areas, and analyze their associated wait times.

Motivations & Goals

We decided that ridesharing apps are a good simulation candidate for two reasons:

1. Ridesharing technology has been developed in the SF Bay Area. Prominent companies such as Uber and Lyft have already optimized their technology, but their process remains hidden from the public (much like Google’s PageRank or Facebook’s News Feed). These algorithms are part of valuable IP, and are highly complex and coveted. So while we have access to the front-end of Uber, the back-end structure remains unknown. Our goal is to simulate the basic process of connecting supply and demand, based on drivers and riders’ location. We won’t be as successful as big tech companies, who have millions of dollars to hire the best engineering talent in the world, but we hope it may provide insight into a service that we frequently use and the crucial decisions one must take to make the system work.

2. Despite the “black box” paradigm, we can still collect large amounts of data from Uber Movement to simulate how their process might work. Uber Movement is an online portal, where anyone can collect data on travel times, travel speeds and mobility heatmaps. Data is available for San Francisco, allowing us to locally optimize the solution. Data can be exported in CSV format and is available under a Creative Commons license, allowing permission of use for academic research. We can also reason from first principles and compare how well our model runs vs. Uber or Lyft, and tweak our model along the way.

In short: we want to simulate a technology that’s popular in our area and we use frequently or have used in the past, but whose inner workings remain a mystery to the public. We hope that simulating ridesharing provides insight into the crucial parameters that make the system work, allowing us to “look behind the curtain”.

Potential Benefits

If we are successful, we might be able to generalize and extrapolate our supply & demand matching algorithm to other use cases. We could release our model to an open-source community, so other people can use our simulation as a starting point for their own model. Depending on the capabilities of SIGMA, we might even find a way to optimize local traffic, based on our model and the Uber Movement data.

Building the model will also teach us which parameters are crucial to develop a successful simulation, and provide insight into the engineering decisions that companies must make to create a good product. These insights will prove invaluable when we apply for the job market.

Qualifications

Brent is interested in automation, machine learning, mobile technology and finance. Ridesharing apps combine these technologies to provide a service to society. Furthermore, he wants to improve his data analysis skills and understand which factors play a decisive role in deploying simulation models and in designing complex systems with many moving parts.

Zain is interested in information technology, automation, and supply chain. Ridesharing platforms elicit different combinations of these. My goal is to provide a better quality service to the good of consumers with the publication of this model. More importantly my intent is to attain as well as showcase my analysis skills with this given project.

Elfy is interested in the social impact of ridesharing. Taxis have dramatically reduced in popularity and there’s a large number of young adults that have begun to reconsider their intent of learning to drive. There are social, financial, and environmental issues to consider with regards to ridesharing applications, aside from just the practical use after a party or visiting local places with limited parking.

Group 3: Members

Brent Van Kersavond Elfy Arrizon Zain Mirza