Embeddings for Flight Price Prediction. The dataset has few variables (csv file), being a true challenge to improve the model's performance. Here, we test the quality & robustness of embeddings for categorical and numerical variables. Also, we evaluate the performance metrics of the model when we implement feature engineering.
We built a robust neural network, conducted feature engineering and transformations, evaluated regression and classification performance, and thoroughly analyzed embeddings using advanced visualization techniques (t-SNE, PCA, dendrograms) to validate model accuracy and representation quality.
Embedding visualizations (t-SNE, UMAP, dendrograms) reveal the internal structure, quality, and meaningfulness of learned categorical representations, guiding model improvements and enhancing interpretability.