DubAir - AirBnB Price Analysis for Dublin, Ireland (Deep Learning Seminar, University of Göttingen, 2022)

This Repo contains several files and folders. In the following a short description is given for the most relevant parts. In general - the files in the main directory contain the most relevant code for the analysis (Data processing, Models,...) The "load_data.py" script contains the class which handles the data loading process with several transformations.

Notice: Notebooks for the image models are not productive as the image data exceeds the capacity of this repo

Folders:

• TabNet_selected - gridsearch, model, histories
• TabNet_correlated - gridsearch, model, histories
• archive - files not used anymore for several reasons
• data - originally downloaded data and generated image variables
• img_models - further folders for different image models with saved models, checkpoints, grid search results etc.
• munich - notebooks and csv's concerning munich prediction
• plots - Plots mainly used for Paper
• logs - Several Tensorboard logs to track model training and checkpoints to restore models
• text_data - saved csv's used for load_data.py as well needed Files within Text_Analysis.ipynb
• xgBoost_models - results and models

Data generating and similar:

• Imputation_New.ipynb (looking insight variables where imputations were needed)
• OSM.ipynb (work on spatial data)
• Text_Analysis.ipynb (work on text variables)
• image_scraping.ipynb (scraping images - adapted for Munich/Dublin)

Feature Selection:

Decorrelate Variables by visual inspection

• feature_selection.ipynb

Models including grid searches:

• CompoundModel.ipynb
• RoomModel.ipynb
• TabNet_GS_fold_Shap.ipynb (uncorrelated features)
• TabNet_GS_fold_Shap_all.ipynb (correlated features)
• ImageModel.ipynb
• xgBoost.ipynb

File holding interim results

• StreetData.csv

Helper files:

• PCA.ipynb (used for experimenting with PCAs during feature_selection.ipynb)
• helpers.py (functions used in several scripts)
• helpers_image_classification.py (functions for image handling in RoomModel.ipynb and image_price.ipynb)
• sentiment_dictionary.csv (NLTK, for Text_Analysis.ipynb)

Content-making for Paper

• plot_tables_paper.ipynb

Technical folders or files

.idea, TabNet_GS_logs, pychache, .DS_Store and .gitignore