How mikesh112/real-estate-ml-econ works
Purpose
Notebook-based Russian real-estate pricing analysis project. It preprocesses CIAN housing data, explores price drivers, and compares linear-regression and ensemble ML approaches for interpretable property valuation.
Entry Points
• EDA.ipynb — primary starting point for raw-data preprocessing and exploratory data analysis.
• ML_Econ_Real_Estate.ipynb — linear-regression modeling workflow.
• ML_Econ_Real_Estate_2.ipynb — ensemble-model workflow.
• Interpretiruemoe-cenoobrazovanie-nedvizhimosti.pdf — presentation of the project’s valuation/interpretability conclusions.
Data Asset and Boundary
• cian_processed_data-2.csv is the processed dataset consumed by modeling notebooks.
• The repository does not expose an application, API, package, or deployment runtime; execution occurs interactively in Jupyter notebooks.
• The raw source-acquisition process is not represented in the tree. The durable handoff boundary is from EDA.ipynb to cian_processed_data-2.csv.
Stage 1: Preparation and EDA
EDA.ipynb performs primary data preparation and exploratory analysis. Its architectural responsibility is converting initial housing observations into a model-ready tabular dataset, identifying feature distributions, quality issues, and relationships relevant to price prediction.
Grounding paths:
• EDA.ipynb
• cian_processed_data-2.csv
Stage 2: Linear Econometric Baseline
ML_Econ_Real_Estate.ipynb evaluates linear regression. This is the interpretable baseline: property attributes become explanatory variables and a real-estate price-related target is estimated through a linear model. Results establish a transparent comparison point for more flexible methods.
Grounding path:
• ML_Econ_Real_Estate.ipynb
Stage 3: Ensemble Modeling
ML_Econ_Real_Estate_2.ipynb evaluates ensemble machine-learning methods against the same processed-data domain. This stage captures nonlinearities and feature interactions that the linear baseline may miss, trading some direct interpretability for predictive capacity.
Grounding path:
• ML_Econ_Real_Estate_2.ipynb
Primary Flow and Outputs
1. Housing data is prepared and examined in EDA.ipynb.
2. Processed records are persisted as cian_processed_data-2.csv.
3. The linear notebook loads/uses processed data for baseline valuation analysis.
4. The ensemble notebook uses the processed-data domain for alternative predictive models.
5. Findings are communicated in Interpretiruemoe-cenoobrazovanie-nedvizhimosti.pdf.
Runtime and Dependencies
The material runtime is Python in Jupyter Notebook/IPython, implied by the .ipynb artifacts. No dependency manifest, service integration, database, cloud infrastructure, scheduled pipeline, or production inference endpoint is present. Reproducibility therefore depends on notebook cell order, locally installed Python ML/data-analysis libraries, and availability of cian_processed_data-2.csv.
Main entry points: cian_processed_data-2.csv Processed CIAN data · ML_Econ_Real_Estate.ipynb Linear regression workflow · ML_Econ_Real_Estate_2.ipynb Ensemble workflow · EDA.ipynb Preparation & EDA
Components 11
Interactive Jupyter workflow
Python notebooks are run manually.
- Preparation & EDA · Cleans housing observations and examines price drivers.
EDA.ipynb - Local Python/Jupyter · Execution depends on notebook order, local libraries, and the CSV.
Valuation modeling
- Linear regression workflow · Interpretable econometric valuation baseline.
ML_Econ_Real_Estate.ipynb - Explanatory variables · Property attributes derived from processed records.
- Linear valuation model · Estimates the price-related target.
- Baseline findings · Transparent reference for comparison.
- Ensemble workflow · Evaluates nonlinear valuation approaches.
ML_Econ_Real_Estate_2.ipynb - Ensemble models · Capture nonlinearities and feature interactions.
- Ensemble findings · Predictive comparison with the linear baseline.
Other components
- Processed CIAN data · Durable model-ready handoff dataset.
cian_processed_data-2.csv - Valuation conclusions · Communicates valuation and interpretability conclusions.
Interpretiruemoe-cenoobrazovanie-nedvizhimosti.pdf
Connections 13
- Local Python/Jupyter → Preparation & EDA: runs
- Preparation & EDA → Processed CIAN data: persists
- Local Python/Jupyter → Linear regression workflow: runs
- Processed CIAN data → Linear regression workflow: loads
- Linear regression workflow → Explanatory variables: derives
- Explanatory variables → Linear valuation model: fits
- Linear valuation model → Baseline findings: produces
- Local Python/Jupyter → Ensemble workflow: runs
- Processed CIAN data → Ensemble workflow: loads
- Ensemble workflow → Ensemble models: evaluates
- Ensemble models → Ensemble findings: produces
- Baseline findings → Valuation conclusions: informs
- Ensemble findings → Valuation conclusions: informs