lukewbecker / clustering-project

Zillow Clustering Project

Author: Luke Becker

Description:

The purpose of this project is to develop a model that is able to predict the log error of Zillow's Zestimate in predicting values in three counties in California, using the Zillow dataset.

In addition, there are these deliverables: 1. Acquire.py, Prep.py, Wrangle.py, and explore.py files to assist in recreating project. 2. Juptyer Notebook with exploration and modeling of the data. 3. Readme with summary of project outline and data dictionary.

Project Goals

• Discover if by using clustering models and algorithmns, I can produce a model that better predicts the log error of Zillow's Zestimate of property values.
• Utilize Clustering techniques to produce visuals and new features for modeling

Project Planning

Initial Questions:

• Does number of bedrooms matter to the log error of the Zestimate?
• Does location within the state or county affect overall log error?
• Are there clusters of related features, such as location information that can yield more useful derived features to model on?
• Do these features created from clustering actually produce a better model to predict the Zestimate's log error?

Hypothesis Tests:

Initial ideas for testing.

Features: latitude and logitude

• H0: The population means for the 4 location clusters are all equal (no significant difference)

• Ha: The population means for the 4 location clusters are not equal (there is a significant difference)

Features: bedroomcnt and bathroomct

• H0: The population means for the 4 rooms clusters are all equal (no significant difference)

• Ha: The population means for the 4 rooms clusters are not equal (there is a significant difference)

Data Dictionary

Feature	Definition
bathroomcnt	Number of bathrooms in property (includes half bathrooms)
bedroomcnt	Number of bedrooms in property
calculatedbathnbr	Number of both bedrooms and bathrooms in property
calculatedfinishedsquarefeet	Total Square Footage of the property
fullbathcnt	Number of full bathrooms in property (excludes half bathrooms)
centroid_latitude	The centroid of latitude created via clustering
centroid_longitude	The centroid of longitutde created via clustering
age	Number of years since house was built
location_cluster_0	Subset of the location clustering
fips	A federal code designating areas of the country that functions similarly to a zip code

Target	Definition
logerror	Log error of Zestimate vs actual property value