README – Repository for the Expert–Machine Prediction Experiments

This repository contains materials used in the experiments reported in the accompanying manuscript. It provides complete transparency regarding results, experimental workflows, and model evaluation.

Repository Contents
1. XGboost.yaml: Version and dependencies of XGBoost

2. Readme.txt: This file

3. Repo_Experiments.zip
  a. Survey Materials (German)

     Questionnaire (HTML) – the complete survey instrument provided to participants.

     Informed Consent (DOCX) – participant information sheet and consent form, also in German.

     Note: Both documents are provided in German, as this was the language of administration during 
     the study.

  b. Raw Experimental Data

     Raw survey results as collected from participants, prior to any cleaning or transformation.
     (e.g., results-surveyfinal.xlsx)

  c. Labels and Model Outputs

     RealLabels_Model_pred.xlsx – combined ground-truth price labels and model predictions used for
     expert–model comparison.

  d. Training Data (Without Transaction Information)

     Train_Geo_Data.csv – the feature set used for model training and evaluation, with all 
     transaction-related attributes removed to comply with data-protection constraints.

  e. Model Script

     ModelTrain.py – full training script used to generate machine-learning predictions, including:

     - data loading and preprocessing

     - model configuration (XGBoost)

     - cross-validation metrics

     - output generation

Notes

- No personally identifiable information is included in any dataset.

- All Transaction-specific data have been removed before publication.

- All survey and consent materials follow the approved ethics protocol of TU Wien.