MSDS 422: Assignment 2 Report
Author: Fares Abanmy
Mar 10 2025
1
The focus of this project involved studying housing data while developing SalePrice
predictions from exploratory data analysis and feature engineering solutions and regression
modeling practices. The assessment starts with SalePrice distribution (Figure 1) which displays
right-skewed features common in houses statistics. The three numerical attributes LotArea,
YearBuilt and TotalBsmtSF show different magnitudes as well as effects on SalePrice so data
normalization becomes essential before further investigation can continue.
Figure 1: Plot showing the distribution of sales price.
2
Missing data is addressed by imputing numerical columns with their mean and dropping
categorical features with excessive nulls. One-hot encoding ensures compatibility with
regression models. A new engineered feature, TotalSF, combines basement, first, and
second-floor square footage, showing a strong positive correlation with SalePrice.
3
The evaluation includes training of Linear Regression model and Ridge model through
5-fold cross-validation. The model with Ridge performs better than Linear Regression model due
to its lower RMSE figures which implies superior hypothesis generalization. The use of
polynomial and indicator and dichotomous features augments model complexity yet teams with
additional challenges for recognizing overfitting patterns. Through this notebook we focus on
making preprocessing resistant to errors and choosing features with correlation analysis to
evaluate our model performance. Thoughtful feature engineering and regularization practices
show their value in predictive modeling through these findings.
Management/Research Question
The predictive abilities of linear regression and ridge regression models for home prices
are affected by three different methods of feature engineering including generating polynomial
features and using indicator variables and age-linked variables. Which model shows the most
optimal results when applying these modifications?
References
Google. 2025. "Gemini Language Model via Google Colab." Google AI. Accessed March 10,
2025. https://colab.research.google.com.
Kaggle. House Prices Dataset. Accessed February 28, 2025.