KC House Dataset Exploratory Analysis
Analysing the factors that drive property prices
- EDA
- Method
- Python
- Tooling
- Açık kaynak
- Sharing
Overview
"In real estate, everything is location" is true but incomplete. This analysis asks precisely: apart from location, what actually determines price, and how large is location's effect in numbers?
Work on the King County dataset started with a data quality audit: missing values, outliers and illogical records were cleaned. Then univariate distributions, pairwise relationships and the correlation structure were examined. The relationship between price and variables such as square footage, room count, year built, renovation status and grade was visualised separately for each.
The most striking finding is that raw square footage alone is misleading. The relationship between area and price is not linear; past a certain threshold, additional square footage contributes markedly less to price. Similarly, the grade score carries a strong effect on price independent of area. The findings were shared on GitHub as a visualised notebook.
Highlights
- Data quality audit and outlier cleaning
- Distribution and correlation analysis per variable
- Showing the area-price relationship is not linear
- An annotated, re-runnable notebook
Objectives
- Establish statistically which variables determine price
- Test intuitive assumptions against the data
- Document findings reproducibly
Solution
- A systematic exploratory data analysis flow
- Testing relationships through visualization
- Identifying non-linear relationships
- Justifying findings inside the notebook
Approach
- 01Auditing data quality
- 02Univariate and multivariate analysis
- 03Visualising and interpreting the findings
Outcomes
- The most influential variables on price were ranked numerically
- The diminishing return of square footage was shown clearly
- The analysis became a template adaptable to other datasets
Challenges
- Cleaning outliers without distorting the result
- Not confusing correlation with causation
- Noticing non-linear relationships
Selected Work
Similar projects
Competitor Store Analysis Platform
A competitor analysis platform for e-commerce businesses, bringing price comparison, assortment breadth and trend tracking into one decision support screen.
Word Cloud Generator
A web application that generates customisable word clouds from text data, combining frequency analysis with visualization options.
Beauty Salon Appointment System
A system that brings appointments, staff and client records for beauty salons into one panel. Clients book online around the clock while the business manages everything from a single screen.