I use machine learning, specifically a basic neural network with LSTM and multi-head attention layers to predict glassdoor job review ratings. Additionally, I use GloVe embeddings, TF-IDF and LDA to preprocess and analyze the text data.
Based on the real data, I achieved 15th out of 45 students in the prediction competition with an R squared of 0.635 in the test data.
Please reach out via LinkedIn if you would like the training data.