https://github.com/dmarks84/ind_project_california-housing-data--kaggle

Independent Project - Kaggle Dataset-- I worked on the California Housing dataset, performing data cleaning and preparation; exploratory data analysis; feature engineering; regression model buildings; model evaluation.
https://github.com/dmarks84/ind_project_california-housing-data--kaggle

cross-validation data-modeling data-reporting data-visualization eda folium grid-search matplotlib model-evaluation numpy pandas pca python seaborn sklearn statistics supervised-ml unsupervised-ml

Last synced: about 6 hours ago
JSON representation

Independent Project - Kaggle Dataset-- I worked on the California Housing dataset, performing data cleaning and preparation; exploratory data analysis; feature engineering; regression model buildings; model evaluation.

• Host: GitHub
• URL: https://github.com/dmarks84/ind_project_california-housing-data--kaggle
• Owner: dmarks84
• License: bsd-3-clause
• Created: 2024-01-30T23:53:22.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2024-03-28T22:14:50.000Z (7 months ago)
• Last Synced: 2024-03-29T22:40:55.648Z (7 months ago)
• Topics: cross-validation, data-modeling, data-reporting, data-visualization, eda, folium, grid-search, matplotlib, model-evaluation, numpy, pandas, pca, python, seaborn, sklearn, statistics, supervised-ml, unsupervised-ml
• Language: Jupyter Notebook
• Homepage: https://www.kaggle.com/code/danpmarks/california-housing-prices-1990
• Size: 2.02 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Ind_Project_California-Housing-Data--Kaggle

    

## Screenshot

![screenshot](https://github.com/dmarks84/Ind_Project_California-Housing-Data--Kaggle/blob/main/housing_screenshot.png?raw=true)

## Summary

I worked on the California Housing Data dataset.  The dataset provided 9 features that I used to try to predict the median housing data.  I performed expanded exploratory data analysis in order to consider the best ways to encode categorical data and scale the numeric features.  I also used PCA to add features, as well as K-means clustering to identify potential meaningful cluster identifiers.  THis informaiton was fed into several regression models/estimators, all of which were then run through gridsearch and cross validation to determine the best model on split training and validation data.  

## Results

### Model Selected & Scoring

Of the models considered, the best was was a **Gradient Boosted Decision Tree with max depth set to 8** (with default parameters otherwise [e.g., learning rate = 0.1]).  The score on validation data was **~0.84**, which is a respectable score but there is likely room for improvement.  It scored well on the training data (0.95), so it is unlikely that the model is overfitting.

## Skills (Developed & Applied)

Programming, Python, Statistics, Numpy, Pandas, Matplotlib, Scikit-learn, Dataframes, Data Modeling, EDA, Data Visualization, Data Reporting, Classification, Supervised ML, Cross Validation, Grid search, Unsupervised ML, PCA, Seaborn, Folium