My first data engineering project, creating a Synapse datawarehouse
Some of the data is from Divvy, a bike sharing program in Chicago, Illinois. This data is anonymous, so Udacity provided some fake rider and account profiles to go along with fake payment data.
A rider can purchase a pass at a kiosk or use a mobile application to unlock a bike at stations around the city and use the bike for a specified amount of time. The bikes can be resturned to the same station or to another station.
Using the Azure tech stack, I wanted to create the following architecture:
In Azure, I needed to create a Postgres database and loaded the original data into it.
Then, I performed some exploratory data analysis (eda) using python to get a look at the original datasets.
Next, the PostgreSQL database was linked to Azure blob storage and the data was loaded as external tables to SQL database in Azure Synapse Analytics.
During the transfer to Synapse, the data was transformed from a relational datamodel, to a star schema.
The original model:
The newly created star schema:
Analyze how much time is spent per ride:
- Based on date and time factors such as day of week and time of day
- Based on which station is the starting and / or ending station
- Based on age of the rider at time of the ride
- Based on whether the rider is a member or a casual rider
Analyze how much money is spent:
- Per month, quarter, year
- Per member, based on the age of the rider at account start
Analyze how much money is spent per member:
- Based on how many rides the rider averages per month
- Based on how many minutes the rider spends on a bike per month