tomerpo5/tomerpomerantz_nyc2019_airbnb_eda

1. 1.Overview

This project presents an exploratory data analysis (EDA) of the Airbnb 2019 dataset for New York City. The objective of the analysis is to explore how price, availability, review activity, and host behavior interact. The project also includes a data-driven identification and analysis of “superhosts.”

1. 1.Dataset Description

Source: gradio/NYC2019-Airbnb-Open-Data (Hugging Face) Observations: 48,895 Features: 16 original variables plus engineered features.

Main Variables: price, room_type, reviews, availability, neighbourhood_group.

1. 1.Data Cleaning

3.1 Removed irrelevant column name,latitude,longitude(not useful for my analysis).

3.2 Handled missing values:

• —host_name: replaced missing values with "Unknown".
• —last_review: replaced missing values with the earliest available date in the column.
• —reviews_per_month: replaced missing values with 0.

3.3 Encoded room_type numerically:

• —Entire home/apt → 1
• —Private room → 2
• —Shared room → 3

3.4 Created new columns:

• —reviews_category: classified listings based on total reviews into ranges (0, 1–5, 6–10, 11–15, 16–20, 20+).
• —is_superhost: data-driven classification of hosts based on reviews, activity, and availability metrics.

3.5 Verified all missing values were handled and column data types were consistent.

1. 1.Exploratory Data Analysis

4.1 Price Distributions Based On Main Variables

Figure 1. Price Distributions

Figure 2. Correlation Matrix

Key findings: Numerical features show limited linear correlation with price. Review activity variables correlate negatively with availability. Variation in price is better explained by categorical features such as room type and location.

4.2 Availability vs. Mean Review Category

Figure 3. Average Availability by Review Category and Borough.

Figure 4. Review Category Counts

Summary: A large proportion of listings have 0–5,20+ reviews, Location is the main effect on avalabilty.

1. 1.Superhost Analysis

5.1 Definition A data-driven approach was used to identify potential superhosts by applying quantile thresholds:

• —Top 25% of numberofreviews.
• —Top 25% of reviewspermonth.
• —Bottom 25% of availability_365.
• —Veiws ≤ 75% in calculatedhostlistings_count.
• —Hosts meeting all criteria were flagged as is_superhost = True.

5.2 Results

• —The characteristics of superhosts show distinct behavioral patterns.
• —Metric Superhosts.
• —Price Slightly higher Moderate.
• —Availabilit is Lower.
• —Higher Review frequency.

Figure 5. Distribution of Superhosts

Summery: Superhosts appear primarily in high-demand boroughs such as Manhattan and Brooklyn.

1. 1.Key Insights
2. 2.Price variation is driven primarily by categorical factors such as room type and borough.
3. 3.Review activity exhibits a negative relationship with availability, serving as an indicator of demand and occupancy.
4. 4.Most listings have limited review activity.
5. 5.Data-driven superhost identification highlights a group of highly active hosts with measurable behavioral patterns: higher reviews, lower availability, and competitive pricing.
6. 6.Superhosts are clustered in central boroughs with high tourist activity.

1. 1.Tools and Libraries Python Pandas NumPy Matplotlib Seaborn Google Colab

1. 1.Author

Tomer Pomerantz Introduction to Data Science Reichman University 2026

Presentation Video - https://www.loom.com/share/dcf270a8bd874453a244b53b403e3a49