Uber uses AI, ml and DS for its major purposes. It is an international multinational ride hailing company offering services that includes peer to peer ridesharing, ride service hailing, food delivery and a micro mobility system with electric bikes and scooters.
Uber was founded by Travis Kalanick, Garrett Camp, Oscar Salazar, Travis Kalanick in 2009 at San Francisco Bay Area, U.S.A.
With total funding of 25.2B USD (2019), the main Investors are SoftBank Vision Fund, Tencent Holdings, Toyota Motor Corporation, and others.
Sub-Organization operate in:
In this article we will see how Uber uses data Analytics in their Business and process and how AI and Data science works for Uber’s Customer support system.
So, let’s see how data analysis is used in Uber’s business process to get growth in business and provide efficient service to customers.
Data analysis of how Uber uses AI
The specific business one that they are looking into is around airport case study that they are working on, so this is a real life problem that they are facing in one of their cities in terms of understanding how the marketplace is held from demand & supply working at airport and to airport.
Data in terms of identifying what could be the problem, could be either from customers tickets or operations real hands on understanding that airport reliability is optimal there are some issues with customers getting trips to the airport right so that’s an unstructured bit of understanding what the problem coming from structure data actually is.
Start with then understanding making out as their datasets most of the data they use for an analytical purpose is stored within their data warehouse & from there they can get the various tables that they can specific trip level information to draw their level of information.
So, this level of information gets them to understand exactly what’s happening at a particular marketplace. How does the daily request come, cancelation rate, rates of completeness, time for completing a particular trip from one destination to another destination etc.
So, the matrices that Uber is looking for specifically completes the request, so this is the matrix that tells you the reliability of how well the market place is.
So specifically, in airport problem they look at different point of the day like from 4 AM to midnight understanding at which hours do they get the request. what proper Percentage of request get completed so, higher the complete ratio means there is better marketplace health in the compare to on lower one’s there is not enough supply to meet the demand.
They started looking at it, they saw cancelation towards the airport is higher compared to the overall cancelation rate for regular trips.
- Airport trips takes long time to get complete and get the next trip
- Amount of time the driver has to wait for another request to get back.
- There is no economic sense.
So, they just figuring out time when the airport traffic demand is high. However, what are the condition for to supply that requirement
And another observation is idle time of drivers at the airport.
- They analyse airport pattern drive idle time inflow for request and wait for new request throughout the day.
- Inflow of cabs to airport
- Outflow of cars to airport at higher demand time
- Time driver have to wait
- Time to request get complete
So, this is all about data analysis used in Uber’s business process.
Let’s move towards the AI and Data science system of Uber customer support process.
Uber uses AI for Customer Process
Uber customer obsession team is responsible for taking care of customers queries & resolving them,
As uber continues to grow on a large scale, support agents must be able to handle an even increasing volume & diversity of customers.
lacks customer tickets created daily on platforms across worldwide.
Uber created a tool called ‘COTA’ (Customer Obsession ticket assistance) which uses Machine learning. Additionally, NLP technique help agents deliver to get better customer support.
- UBER uses Michelangelo (Uber’s own ML service as platform) for quick and efficient resolution for more than 90% of tickets.
-
- It works in english but model in developing to work in spain & portuguses also
- suggests most 3 likely issues types & solutions based on ticket content and trip content.
- It can reduce ticket resolution time by 10% while delivering service with similar or higher levels of customer satisfaction.
Architecture has 7 main steps. The 1st says that once a new ticket enters the customer support platform known as CSP.
the backend service collects all relevant features of the tickets.
2nd the backend service then sends these features to the Machine learning model In Michelangelo.
3rd model is predicting the score for each possible solution
4th backend service receives the prediction & score and says it’s going to scheme the data store.
5th once an agent opens a given ticket the front end service triggers the backend service to check if there are any updates to the ticket if there is no update the back end service will retrieve the saved predictions, if there are any update it will fetch the updated feature and go to the step 2 to 4 again 6 backend service returns the list of solution ranked by predicted score to the front end on the last step the top 3 rank solutions are suggested to the agents.
From their agents make a selection out of these 3 and resolve the user’s problem.
It achieves 2 tasks:-
- Identifying ticket issue type.
- Determining the most accurate solution.
To achieve these two tasks the ML model uses feature extraction customer support message, trip information and Customer selection in the ticket issues.
Build a NLP model pipeline to convert those messages into useful features for the model.
Conclusion | Uber Uses AI
So that’s how uber works. I hope you find this article insightful!
Article By: Sachin Dubey
If you are Interested In Machine Learning You Can Check Machine Learning Internship Program
Also Check Other Technical And Non Technical Internship Programs
