Predictive Thursdays: For All Uber and Lyft Drivers—A Guide to “Driving Less and Making More” Part Two

In my previous blog, we discussed the two very important questions that most Uber drivers would love to know the answers to—”How much will I earn by the end of the week?” And, “Is there a way to crack the code to drive less and make more money?” I discussed how to use Time Series Forecasting techniques to figure out my earnings (see the blog for the models). Now comes the tougher question—How can I maximize what I make? How do I choose “How much,” “When,” and “Where exactly” I should drive?

Graph Theory Combined with Geospatial Analysis

I can choose where I should drive by analyzing the routes that I take with graph theory and geospatial analysis. Sounds complicated, doesn’t it? It isn’t. SAP Predictive Analytics provides you with Social Network Analysis to build a graph using geospatial (latitude & longitude) data. If you have access to your latitude and longitude data, the product can automatically combine these two to generate “Position” as a new variable. The graph analysis gave me a fair idea of the number and duration of each trip.

Now, let me back up a little and take a look at my trips’ data to understand what’s going on. Below are some of the screenshots from my app which show my trips. (Please observe the timing). These trips are either:

1. Between Downtown to Airport
2. Downtown to Downtown (within Downtown)
3. Downtown to Suburbs

Well … As an Uber driver, I have limited options due to business rules. I can’t cancel more than three times, and I never know where my final destination is going to be (I only find this out once the passenger sits in my car and starts the ride). However, after a year of experience, I have a pretty fair idea that if a person is asking for an Uber from 1 Michigan Avenue at 5:00 PM, there is a high chance he’s going somewhere within Downtown vs. if a person is asking for an Uber from a Courtyard Downtown Marriott at 6:45am.

With that in mind, let’s look at some more graphs of my rides.

So, you can clearly observe that it’s not just the source and the destination that are playing a role in my final earnings. There are lots of other variables that could be very important, such as Time of the Day, Day of the Week, Day of the Month, Weekend, Snowing, and so on.

So how do I figure out what is the perfect combination of variables that will give me the maximum revenue? How do I crack the code?

How to Drive Less and Make More

This question translates to a problem where I have to predict a continuous variable (earnings) given different variables (described above). Aha! Since this is a Regression problem, I just go back to my SAP Predictive Analytics and use its Regression wizard and feed all the above data items to create a regression model.

The most critical output from SAP Predictive Analytics is the chart below which explains the influential variables that contribute to my earnings (which is what I wanted to understand). 

1)Which Day and Which Weather Is the Most Profitable?

First of all, this model has helped clear up some of my misconceptions, like my assumption that the more I worked, the more money I would make. Because you can clearly see that “Hours_Worked” is not really the most important variable.

Out of the top five variables, the first three are really indicators. If  I’m working on either the first Friday of the month or any Saturday (weekend) of a snowy day, I’m going to make more money. No Brainer. However, let’s look closer at the Pickup_Zip (by double clicking it).

2)Which Zip Code Is the Most Profitable?

The first two categories are the zip codes of the Downtown and Airport area, which means that’s where I’ll get the maximum revenue. However, if I end up picking someone from the other zip codes, I’m going to negatively impact my revenue potential. That means, I’ll still be driving more, but not making enough money.

3)Which Time of Day Is the Most Profitable?

Looking closer at the Hour_of_the_day variable below, I get a fantastic analysis of the exact hour of the day when I should be working hard and picking up more passengers. The early evenings are going to fetch me the maximum revenue!

WOW! This is fantastic information. I can even go a step further and create a simulation based on all these factors to understand which is that best combination that gives me the maximum earning.

And once I know the best combination(s), I can…

Drive Less and Earn More $$$

Hope you enjoyed reading this! Let me know what you think….

Learn More

• Download “The Forrester Wave: Predictive Analytics And Machine Learning Solutions, Q1 2017”
• For more on machine learning topics, read the other blogs in our Thursday series