Property Value and Predictive Model Building

Project Description

This project requires analyzing and building some models to assess and predict the assessed value of properties in the greater Boston area. After comparing, choose the model with best performance and decide which variables are the top influencers of property value.

Tools and Learning Methods

Decision Tree, Gradient Boosting (XGBoost), k-fold cross validation

Raw Data

Training set of 10k+ records Test set of 4k+ records

Data Dictionary 1

Name	Description
PID	Unique parcel number
ZIPCODE	Zip for the parcel
OWN_OCC	One-character code indicating if owner received residential exemption as an owner-occupied property
AV_TOTAL	Assessed value for property
LAND_SF	Parcel's land area in square feet
YR_BUILT	Year property was built
YR_REMOD	Year property was last remodeled
LIVING_AREA	Living area square footage of the property
NUM_FLOORS	Number of levels in the structure located on the parcel
STRUCTURE_CLASS	Structural classification of commercial building
R_BLDG_STYL	Residential buiilding style
R_ROOF_TYP	Structure roof type
R_EXT_FIN	Structure exterior finish
R_TOTAL_RMS	Total number of rooms in the structure
R_BDRMS	Total number of bedrooms in the structure
R_FULL_BTH	Total number of full baths in the structure
R_HALF_BTH	Total number of half baths in the structure
R_BTH_STYLE	Residential bath style
R_KITCH	Total number of kitchens in the structure
R_KITCH_STYLE	Residential kitchen style
R_HEAT_TYP	Structure heat type
R_AC	Indicates if the structure has air conditioning (A/C)
R_FPLACE	Total number of fireplaces in the structure
R_EXT_CND	Residential exterior condition
R_OVRALL_CND	Residential overall condition
R_INT_CND	Residential interior condition
R_INT_FIN	Residential interior finish
R_VIEW	Residential view

Data Dictionary 2

Name	Description
ZIP	ZIP CODE
POPULATION	Population of people in the ZIP code
POP_DENSITY	People per square mile
MEDIAN_INCOME	Median Income of the residence of that zip code
City_State	City Name and State