Project 2: Ames Housing Data and Kaggle Challenge

---

Problem Statement:

• Creating a regression model based on the Ames Housing Dataset. This model will predict the price of a house at sale.
• The main target audience will be property agents having trouble evaluating the price of each property in ames.

---

Background:

Ames is a city in the state of Iowa of Unite States. In 2020, Ames had a population of about 66,000. Majority of the population are in the 15-30 years age range (roughly 50%). As of 2010, which is in the same period as our project, there were 23,876 housing units. Our model will be based on about 2000 datasets , which is roughly 10% of the total housing units. The average household size was 2.25 and the average family size was 2.82. In 2010, Ames had a real per capita income of $30,128. (source). These numbers give an indication on the types of house built in Ames and how it is priced, which we will look at later. (source)

Another interesting point to note is that the number of housing units outpaced the growth in total households. This could mean that supply is more than demand. However one cannot be sure, especially with data showing that 30% of householders are living alone (source). So the increased supply of housing units could be due to individuals who chose to stay alone. Either way, we will attempt to predict the housing price based on multiple features.

Equipped with this knowledge, we set to achieve a good model prediction for house sales price based on datasets of various house features from 2006-2010.

---

Datasets selected:

• train.csv
• The dataset provides data on 81 different features related to the housing units in Ames. These features are separated between categorical and numerical variables.

Feature	Type	Dataset	Description
Id	object	train.csv	ID of house
MS SubClass	object	train.csv	The building class
MS Zoning	object	train.csv	Identifies the general zoning classification of the sale
Lot Frontage	float64	train.csv	Linear feet of street connected to property
Lot Area	int64	train.csv	Lot size in square feet
Street	object	train.csv	Type of road access to property
Alley	object	train.csv	Type of alley access to property
Lot Shape	object	train.csv	General shape of property
Land Contour	object	train.csv	Flatness of the property
Utilities	object	train.csv	Type of utilities available
Lot Config	object	train.csv	Lot configuration
Land Slope	object	train.csv	Slope of property
Neighborhood	object	train.csv	Physical locations within Ames city limits
Condition 1	object	train.csv	Proximity to main road or railroad
Condition 2	object	train.csv	Proximity to main road or railroad (if a second is present)
Bldg Type	object	train.csv	Type of dwelling
House Style	object	train.csv	Style of dwelling
Overall Qual	int64	train.csv	Overall material and finish quality
Overall Cond	int64	train.csv	Overall condition rating
Year Built	int64	train.csv	Original construction date
Year Remod/Add	int64	train.csv	Remodel date (same as construction date if no remodeling or additions)
Roof Style	object	train.csv	Type of roof
Roof Matl	object	train.csv	Roof material
Exterior 1st	object	train.csv	Exterior covering on house
Exterior 2nd	object	train.csv	Exterior covering on house (if more than one material)
Mas Vnr Type	object	train.csv	Masonry veneer type
Mas Vnr Area	float64	train.csv	Masonry veneer area in square feet
Exter Qual	int64	train.csv	Exterior material quality
Exter Cond	int64	train.csv	Present condition of the material on the exterior
Foundation	object	train.csv	Type of foundation
Bsmt Qual	int64	train.csv	Height of the basement
Bsmt Cond	int64	train.csv	General condition of the basement
Bsmt Exposure	object	train.csv	Walkout or garden level basement walls
BsmtFin Type 1	object	train.csv	Quality of basement finished area
BsmtFin SF 1	float64	train.csv	Type 1 finished square feet
BsmtFin Type 2	object	train.csv	Quality of second finished area (if present)
BsmtFin SF 2	float64	train.csv	Type 2 finished square feet
Bsmt Unf SF	float64	train.csv	Unfinished square feet of basement area
Total Bsmt SF	float64	train.csv	Total square feet of basement area
Heating	object	train.csv	Type of heating
Heating QC	int64	train.csv	Heating quality and condition
Central Air	object	train.csv	Central air conditioning
Electrical	object	train.csv	Electrical system
1st Flr SF	int64	train.csv	First Floor square feet
2nd Flr SF	int64	train.csv	Second floor square feet
Low Qual Fin SF	int64	train.csv	Low quality finished square feet (all floors)
Gr Liv Area	int64	train.csv	Above grade (ground) living area square feet
Bsmt Full Bath	float64	train.csv	Basement full bathrooms
Bsmt Half Bath	float64	train.csv	Basement half bathrooms
FullBath	float64	train.csv	Full bathrooms above grade
HalfBath	float64	train.csv	Half baths above grade
Bedroom AbvGr	int64	train.csv	Number of bedrooms above basement level
Kitchen AbvGr	int64	train.csv	Number of kitchens
Kitchen Qual	int64	train.csv	Kitchen quality
TotRms AbvGrd	int64	train.csv	Total rooms above grade (does not include bathrooms)
Functional	object	train.csv	Home functionality rating
Fireplaces	int64	train.csv	Number of fireplaces
Fireplace Qu	int64	train.csv	Fireplace quality
Garage Type	object	train.csv	Garage location
Garage Yr Blt	float64	train.csv	Year garage was built
Garage Finish	object	train.csv	Interior finish of the garage
Garage Cars	float64	train.csv	Size of garage in car capacity
Garage Area	float64	train.csv	Size of garage in square feet
Garage Qual	int64	train.csv	Garage quality
Garage Cond	int64	train.csv	Garage condition
Paved Drive	object	train.csv	Paved driveway
Wood Deck SF	int64	train.csv	Wood deck area in square feet
Open Porch SF	int64	train.csv	Open porch area in square feet
Enclosed Porch	int64	train.csv	Enclosed porch area in square feet
3Ssn Porch	int64	train.csv	Three season porch area in square feet
Screen Porch	int64	train.csv	Screen porch area in square feet
Pool Area	int64	train.csv	Pool area in square feet
Pool QC	int64	train.csv	Pool quality
Fence	object	train.csv	Fence quality
Misc Feature	object	train.csv	Miscellaneous feature not covered in other categories
Misc Val	int64	train.csv	$Value of miscellaneous feature
Mo Sold	object	train.csv	Month Sold
Yr Sold	object	train.csv	Year Sold
Sale Type	object	train.csv	Type of sale
SalePrice	int64	train.csv	Property price in dollars
combined_living_area	float64	train.csv	total living area
total_bath	float64	train.csv	total bathrooms
age_Year Built	int64	train.csv	number of years since home built
age_Year Remod/Add	int64	train.csv	number of years since home remod/add
age_Garage Yr Blt	int64	train.csv	number of years since garage built
neighborhood_ranking	int64	train.csv	neighborhood ranking based on categories

---

Findings

• Some datas were missing and had to be imputed using several techniques such as imputing the mean value, assigning 0 values, and assigning as none.
• 1 outlier was corrected using median.
• Further Exploratory Data Analysis (EDA) showed multiple need to remove certain features that would adversely affect our model prediction.
• These includes removing one of two features with multicollinearity, removing certain features with very low correlation to SalePrice, and engineering new features to get more insights and achieve a better prediction model.
• Further data preprocessing was required such as removing features that have only 1 value or class dominating.
• 3 models were considered for our datasets, with the Ridge Regression Model coming up as the best prediction model for SalePrice.
• The plot above shows the predicted vs actual value of our test dataset. Our model clearly outperforms the null model (orange line) and is a good predictor of SalePrice based on the features input.
• With an R2 score of 0.855, it means 85.5% of variations can be explained by our model, which is a relatively good predictor.
• In general the chart below shows the top 25 features that affects SalePrice positively and negatively.

---

Conclusions/Recommendations

• In conclusion, our model has provided sufficient information for property sales agent to better price their properties using features selected. The model is able to explain 85% of the variations and that can be considered good for the case of property sales agent.

• It is highly recommended for property sales agent to emphasize on the following positive affecting and negative affecting features in relation to SalePrice.

• POSITIVE FEATURES TO INCLUDE

• Overall material and finish quality

• Neighborhood Ranking (for neighborhoods in the top category 'Timber', 'Blmngtn', 'Veenker', 'Somerst', 'NoRidge', 'StoneBr', 'NridgHt')

• Lot Area (an increase of 100 sq foot leads to increase in SalePrice of approximately $10,000)

• Exterior Covering using Cement Board

• Fireplace Quality

• Total bathrooms (the more the better)

• Built-in garage (the space above the garage seems to be highly valued in Ames, as opposed to a normal attached garage)

• NEGATIVE FEATURES TO AVOID

• Basement Exposure (No exposure in basement fared the worst, Ames residents love their sunlight)

• Number of Kitchen AbvGround

• Exterior Covering using Hard Board

• Exterior Covering using Stucco

• Irregular lot shape