These Kaggle datasets offer a comprehensive analysis of the US real estate market, leveraging data sourced from Redfin via an unofficial API. It contains weekly snapshots stored in CSV files, reflecting the dynamic nature of property listings, prices, and market trends across various states and cities, except for Wyoming, Montana, and North Dakota, and with specific data generation for Texas cities. Notably, the dataset includes a prepared version, USA_clean_unique, which has undergone initial cleaning steps as outlined in the thesis. These datasets were part of my thesis; other two countries were France and UK.

These steps include:

• Removal of irrelevant features for statistical analysis.
• Renaming variables for consistency across international datasets.
• Adjustment of variable value ranges for a more refined analysis.

Unique aspects such as Redfin’s “hot” label algorithm, property search status, and detailed categorizations of property types (e.g., single-family residences, condominiums/co-ops, multi-family homes, townhouses) provide deep insights into the market. Additionally, external factors like interest rates, stock market volatility, unemployment rates, and crime rates have been integrated to enrich the dataset and offer a multifaceted view of the real estate market's drivers.

The USA_clean_unique dataset represents a key step before data normalization/trimming, containing variables both in their raw form and categorized based on predefined criteria, such as property size, year of construction, and number of bathrooms/bedrooms. This structured approach aims to capture the non-linear relationships between various features and property prices, enhancing the dataset's utility for predictive modeling and market analysis.

See columns from USA_clean_unique.csv and my Thesis (Table 2.8) for exact column descriptions.

Table 2.4 and Section 2.2.3, which I refer to in the column descriptions, can be found in my thesis; see University Library. Click on Online Access->Hlavni prace.

If you want to continue generating datasets yourself, see my [Github Repository] (https://github.com/ArturDragunov/Master_Thesis/tree/main) for code inspiration.

Let me know if you want to see how I got from raw data to USA_clean_unique.csv. Multiple steps include cleaning in Tableau Prep and R, downloading and merging external variables to the dataset, removing duplicates, and renaming columns for consistency.