Job Recommendation Platform
Published:
About
The goal was to implement machine learning agents to recommend jobs in a simulated LinkedIn environment. We tried standard Machine Learning algorithms shuch as regressions, SVMs, … as well as neural networks.
The results showed that Deep Learning and Linear Regression agents don’t perform well on the environment. But the SVM, Gradient boosting tree, embedding (when the environment is not evolutive) and the dot agents perform really well. Because the reward is partly based on a dot product, the embedding and dot agents results are not surprising, but the SVM and gradient boosting results were better than expected.
Code Link
Stack
TensorFlow, Scikit-Learn