Predicting Boston house prices in RStudio
- Complete analysis can be viewed in RPubs at https://rpubs.com/brosnahj/Boston-House-Predictions
- Complete analysis can also be viewed in Predict-House-Prices.md file above
We have been asked to investigate the Boston House Price dataset. Each record in the database describes a Boston suburb or town.
Can a model be built to predict house prices in Boston Area with 80% accuracy level?
The data was drawn from the Boston Standard Metropolitan Statistical Area (SMSA) in 1970 from UCI Machine Learning Library.
- Comprehensive univariate and multivariate plots
- Linear algorithms: linear regression, logistic regression
- Non-linear algorithms: Support Vector Machines (radial basis), CART, KNN
- Ensemble algorithms: Stochastic Gradient Boosting, Random Forest, Cubist
- Resample accuracy comparison and plots
- Model tuning, grid search
- Business problem solved
House prices can be predicted with 90% accuracy using top Cubist tuned algorithm. The error rate of our top model is plus or minus $3.2 thousand. Business objective has been achieved.
