Hire Heroes

Overview

Hire Heroes USA is a 501c3 nonprofit organization that empowers U.S. military members, veterans and military spouses to succeed in the workforce. They have helped over 30,000 clients find job placement through a number of services. In order to ensure the highest rate of success, machine learning was leveraged in order to recommend offered services based on demographic information of the client.

Approach

Traditional linear models, while offering excellent interpretability, assume that each part is independent of each other part. Whereas, in reality, something like a client’s willingness to travel a longer distance would be more significant if they had more dependents or had more severe disabilities. Additionally, the services provided are not isolated either. Receiving help with interviews and resumes will have a stronger effect when combined with creating a Linkedin account. In order to capture these complex interactions, non-linear, ensemble model structures were selected. This means that the precise, direct correlation between the model feature and the prediction cannot be defined, but the relative information gained from that feature can be calculated.

Dashboard

Prediction View

Through an interactive dashboard, users are able to quickly identify the best combination of services provided by Hire Heroes in order to either maximize the probability of being hired or the minimize the time between starting the search and finding a job. This was accomplished through exhaustive grid search that runs hundreds of thousands of combinations through the models and selects the optimized set of services.

Client View

Users can also review existing client’s information with recommendations for which services would aid them the most and the percentage of improvement in probability.

Service Analysis

More in-depth analysis of each service for that specific client can also be viewed in order to see the total isolated effect of that service in all combinations with other services