Machine learning analysis script

[moniquewong]

In response to this pull request

In analysis.ipynb:

• Separated code that will be used in the script versus code to get us going within Jupyter notebook
• Created a pipeline for preprocessing
• Removed logistic regression as a model to try (this is a classifier I believe)
• Trained 4 models and discounted R^2 as an appropriate scoring mechanism (see unused code) and chose MSE as scoring mechanism
• Created separate training and validation sets within training data for above
• Chose RandomForestRegressor()
• Tuned hyperparameters on RandomForestRegressor() using RandomizedSearchCV()
• Ended up doing the following instead: shortlisted linear regression, svr and knn to do hyperparameter tuning
• Found best hyperparameters for svr and knn
• Checked best versions of svr and knn against linear regression
• chose linear regression
• fit linear regression
• Created plot with residuals

To be done (need to have):

• [x] Fit and tune SVR()
• [x] Create plot for hyper parameter tuning
• [x] Ensure this analysis.ipynb works with wrangling script
• [x] Turn analysis.ipynb into script that outputs model selection table, hyper parameter tuning chart, residuals chart
• [x] Add docstrings for supporting functions
• [ ] Write tests for supporting functions

To be done (nice to have):

• [ ] Feature engineer neighborhoods and property types

Reflection:

• Could have probably experimented with change of basis for LinearRegression()

[moniquewong]

Added analyze_data.py which:

• Takes code from analysis.ipynb to create output figures
• Refactored code into functions to call in main

Added docstrings and tests needed for supporting functions to above to-do list