The car rental company from the Predictive Maintenance use case wants to expand their network in order to satisfy new demand from a partnership with an insurance company. To get maximum value from this expansion, they have approached our data team once again. Their goals include being able to:
Understand how the current network fits with the observed and anticipated demand
Optimize the vehicle rotation schedule by predicting the demand at each agency location
Evaluate partnerships and/or acquisition locations to expand the network efficiently
This use case requires the following three input data sources, available as downloadable archives at the links below:
Demand: A record of all road accidents handled by the insurance company spanning four years. The four yearly files total approximately 250k observations. This dataset is a modified version of a French open dataset.
Network: A simulated dataset that records information on the current locations of nearly 350 car rental agencies.
Partners: A simulated dataset that contains information on potential partners; that is, new locations (garages) to expand the network. The data must be parsed from raw html in order to extract the relevant information.
The final Dataiku DSS pipeline is pictured below. You can also follow along with the completed project in the Dataiku gallery.
The Flow has the following high-level steps:
Upload the datasets
Clean the datasets
Join the datasets based on geographic proximity
Aggregate the data by geography and station
You should be familiar with:
The Basics courses
The Window recipe
The Reverse Geocoding / Admin Maps plugin is required to create the filled administrative map chart.
The Geocoder plugin is required to enrich the demand dataset and identify the geographic coordinates of rental agencies from addresses.
For general notes on plugins to Dataiku DSS, please see the reference documentation.