Real Estate Data Entry & Standardization

A portfolio project demonstrating professional data entry cleanup and standardization on a messy real estate listings dataset — the kind of work done before importing data into a CRM, MLS platform, or database.

---

Project Description

Real estate data often arrives from multiple agents in different formats: prices written as "$200K" or "1.2M", property types spelled 15 different ways, states written as full names or abbreviations. This project takes 200 raw listings and applies a documented, repeatable cleaning pipeline.

---

Screenshots

(Add notebook screenshots here after running the project)

Before	After

Charts generated:

Missing Values	Property Type Standardization

---

Data Problems Solved

Problem	Example	Fix Applied
Duplicate listings	20 exact copies	drop_duplicates()
Missing values	NaN in bedrooms, sqft, price	Median / mode imputation
Property type variants	"house", "HOUSE", "SFH", "single family"	Lookup map -> 5 canonical categories
Mixed price formats	"$200,000", "200K", "1.2M"	Strip symbols, expand K/M, cast to float
Inconsistent city names	"los angeles", "L.A.", "LA"	Lookup map -> canonical name
Inconsistent states	"New York", "N.Y.", "ny"	Lookup map -> 2-letter USPS code
Address formatting	"123 MAIN STREET", "123 main st."	Title Case + abbreviation map
Mixed date formats	"01/15/2023", "January 15, 2023"	pd.to_datetime(format='mixed')

---

Tools Used

• Python 3.9+
• Pandas — data loading, transformation, and export
• NumPy — numeric operations and NaN handling
• Matplotlib — before/after visualizations
• openpyxl — writing the .xlsx output

---

Project Structure

real-estate-data-entry/
├── data/
│   ├── dirty_real_estate.csv        <- generated by generate_dirty_data.py
│   ├── clean_real_estate.csv        <- output of the cleaning pipeline
│   ├── clean_real_estate.xlsx       <- same data as Excel
│   ├── chart_nulls.png              <- missing values chart
│   └── chart_property_types.png     <- property type standardization chart
├── generate_dirty_data.py           <- creates the messy dataset
├── clean_data.py                    <- standalone cleaning script (commented)
├── build_notebook.py                <- generates portfolio.ipynb
├── portfolio.ipynb                  <- full interactive walkthrough
└── README.md

---

How to Run

1. Install dependencies

pip install pandas numpy matplotlib openpyxl notebook

2. Generate the dirty dataset

python generate_dirty_data.py

3. Option A — Run the standalone script

python clean_data.py

3. Option B — Run the interactive notebook

python build_notebook.py        # generate the notebook file
jupyter notebook portfolio.ipynb
# then: Cell -> Run All

---

Results

After running the full cleaning pipeline:

Metric	Before	After
Total rows	200	~175
Null values	82	0
Duplicate listings	20	0
Property type variants	15+	5 canonical
Price data type	mixed strings	float64
State format	mixed (full name / abbr / lowercase)	2-letter USPS code
Date data type	mixed strings	datetime64[ns]

Key actions:

• Removed 20 duplicate listings
• Imputed or dropped 82 null values
• Reduced 15+ property type spellings to 5 clean categories
• Parsed 3 different date formats into a single datetime column
• Standardized all price strings to float64
• Mapped all state variants to the 2-letter USPS code

---

About

Built as part of a data analytics freelance portfolio. Tools: Python - Pandas - NumPy - Matplotlib - openpyxl

Feel free to fork this project and adapt it to your own dataset.