In this analysis, we take a look at the Seattle Airbnb datasets and try to answer the following questions with both exploratory data analysis (EDA) techniques and machine learning algorithms:
- What patterns of pricing variation can be observed?
- Do superhosts charge a premium compared with regular hosts?
- How well can we predict the price, using the given data?
- How well can we predict the host ratings?
Below language, libraries, and packages need to be installed in order to run the code properly:
- Python
- Jupyter Notebook
- Numpy
- Pandas
- Matplotlib
- Seaborn
- Scikit-learn
Three datasets were utilized for the analysis. The datasets can be downloaded from Kaggle.
The three datasets are:
- Listings: This dataset is what we focused on
- Calendar: We think this is the 365-day availability calendar
- Reviews: This dataset provides the detailed reviews in text descriptions
Below are some of the plotting techniques applied to understand the dataset and prediction results:
- Histogram
- Line plot
- Scatter plot
- Box plot
- Heatmap
A summary of our conclusions are:
- Superhosts don't charge more
- The variables we selected have a 0.6 R-squared score when using a linear regression model to predict price
- Host ratings seem more difficult to prediction
For more details, visit our blog here.