4nuG / predicting-house-prices

We thoroughly analyze real estate prices, exploring the main factors using advanced statistics and modeling techniques.

Table of Contents

1. Project Overview
2. Key Steps
   • Data Wrangling
   • Exploratory Data Analysis (EDA)
   • Predictive Model Building
   • Model Validation
   • Conclusion
3. Full Project Write-Up

Project Overview:

House price estimator for a house dataset using JMP.

Datasets

Datasets are called AmesHousing Data Training_Validation.jmp and AmesHousing Data Scoring.jmp and are in the repo.

Requirements

• JMP Pro 16

Key Steps

Data Wrangling:

• Outlier analysis was performed using continuous variables.
• Outliers at specific rows were identified and removed for analysis.
• Missing data analysis revealed no significant issues, and a principal component (PC) analysis was conducted.

Exploratory Data Analysis (EDA):

• PC analysis identified key predictors, providing insights into the dataset structure.
• Exploration of PC1 and PC2 highlighted predictors related to living area and rooms above ground.

Predictive Model Building:

• A standard least squares model was initially created.
• After addressing outliers, a refined model achieved an improved R-square value of 0.9509.
• Additional models, including stepwise and lasso, were employed and compared.

Model Validation:

• Validation processes, such as Max Validation RSquare and model comparisons, were executed.
• Lasso model emerged as the best fit for scoring data, achieving a validation RSquare of 0.9087.

Conclusion:

• The standard least squares model excelled for training data (RSquare: 0.9509).
• The stepwise Max RSquare model performed best for validation data (RSquare: 0.9238).
• When applied to scoring data, the lasso model exhibited superior performance (RSquare: 0.9087).
• Identified predictors from the lasso model were crucial for accurate predictions.

Full Project Write-Up

For a detailed exploration and findings, refer to the full project write-up in the repository.