DMP Poodles vs. Pugs as Income Indicators

The datasets used in this machine learning experiment, Poodles vs. Pugs as Income Indicators, are publicly available through the Open Government Data portal of the City of Vienna (data.gv.at). This project utilizes two specific datasets: Hunderassen Wien and Durchschnittliches Nettoeinkommen seit 2002 - Bezirke Wien. The rows in the processed dataset represent Vienna's 23 municipal districts. 

Context and Methodology

The purpose of this dataset is to train and evaluate a regression model that predicts the average annual income of a geographic area based on the prevalence of specific dog breeds.

The dataset is not newly collected. It is reused from open data sources (data.gv.at). The project workflow involves:

• Acquiring and downloading the raw datasets

• Cleaning and preprocessing the data

• Merging the demographic and income data by municipal district

• Training a machine learning model (Random Forest regressor)

• Evaluating model accuracy and generating visual distributions

Additional datasets and artifacts are created during the project, including a merged/cleaned dataset, model evaluation metrics, and geographic distribution plots.

Technical Details

The dataset and project environment are structured in a clear folder hierarchy:

• data

• results

Files include:

• datasets (CSV): Original downloaded files from Stadt Wien.

• plots (PNG): Geographic distribution of dog breeds (avg_dogs_per_district.png), income distribution (avg_income_per_district.png), and model error rates (final_evaluation_plots.png).

• source code (IPYNB): Jupyter Notebook containing the data orchestration and modeling pipeline.

Software requirements: To open and work with this dataset and its accompanying code, the following open-source tools are required:

• Python 3

• pandas (for data orchestration)

• scikit-learn (for predictive modeling)

• matplotlib & seaborn (for visualization)

• Jupyter Notebook environment

Further Details

The dataset contains no personal or sensitive data. All records are aggregated at the district level, ensuring complete anonymity and compliance with open-data standards.

Important points for reuse:

• The Random Forest model demonstrated robust predictive power for lower-to-middle income brackets.

• The model's accuracy plateaus at the highest income tiers, indicating that "luxury" wealth is influenced by factors beyond simple pet ownership trends. Future researchers should account for this limitation when utilizing the model for high-income predictions.