|Date Posted||November 13, 2019|
|Industry||Newspapers / Wires|
The Bay Area News Group is seeking a machine learning specialist proficient with the Python programming language as well as Sci-Kit Learn packages including Random Forest, Numpy, Pandas and Matplotlib to build a model that scores website visitors by their propensity to purchase a digital subscription. The overall goal of the project is to estimate the value of local news to residents who are closest to where the news happens. The specialist should be comfortable working in Jupyter Notebook and comfortable utilizing Python API documentation provided by our customer data platform.
The position is temporary. The specialist will work with the Bay Area News Group’s digital subscriptions team to support a Google News Initiative Innovation Challenge project.
The specialist will be required to:
Attend several webcalls to discuss database schemas, their set up, and required data flows, joins, and other manipulations to get the data into the shape it needs to be for the model to run. Such calls will likely take an hour once every two weeks in the initial setup phase and could last several months.
Advise on the capture, storage and manipulation of location-based data, including the distance between a geo-tagged story and a person who has allowed their location to be shared with us. Data related to proximity shall be used in the propensity model.
Create a model using Python that takes multiple columns of data into account in creating the score, including pageviews, visits, recency, frequency, number of sections visited, sections, average distance to stories. etc. The scores shall be written to profiles on our customer data platform.
Visualize the importance of different features in the model, including the importance of proximity of clicked stories to the user.
Help visualize the geographic blast radius of stories in terms of their popularity and who clicked on them, with an eye to helping inform the newsroom which among a certain selection of stories had a greater or lesser blast radius, or area of impact.
Help visualize the interest radius of users and subscribers in terms of which stories they clicked on.
Help explain how the models and visualizations work and possibly participate in writing or speaking about the models.
Term: This contract envisions the project to be complete over the course of one year, wrapping up in October 2020. Further work would be by mutual agreement.
Please contact Ryan Nakashima at email@example.com.