Immo_Eliza

Data Analysis

Repository: challenge-data-analysis
Type of Challenge: Consolidation
Duration: 3 days
Deadline: 22/11/2024 12:30
Team Size: 3

Table of Contents

1. The Mission
2. Installation
3. Usage
4. Deliverables
5. Team

The Mission

We are continuing our journey to create a machine learning model that predicts property prices in Belgium. In this phase, we focus on conducting a preliminary data analysis to gather meaningful insights.
We will use the previously scraped dataset, clean the data, and perform the analysis.

Installation

Project Directory Structure

immo-eliza-analysis/
├── assets/                                     # Folder containing files for graphs visualizations
├── Data/                                      # Folder containing all datasets
├── map/                                       # Folder containing files for map visualizations
│   ├── immoweb_data.csv                       # Raw dataset
│   ├── immoweb_data_cleaned.csv               # Cleaned dataset
│   ├── immoweb_data_filtred.csv               # Filtered dataset
│   ├── encoded_data.csv                       # Encoded dataset
├── dashboard.py                               # Build and run dashboard
├── Data_Analysis.py                           # Run analysis on the cleaned data
├── Data_Cleaning.py                           # Run cleaning on raw immoweb_data.csv
├── README.md                                  # Project overview and instructions
├── real_estate_data_analysis.ipynb            # Jupyter notebook for data analysis
├── real_estate_data_cleaning.ipynb            # Jupyter notebook for data cleaning

Clone the Repository

To get started, clone the repository:

git clone https://github.com/ManelBouba/immo_eliza_analysis

Usage

To run the project, follow the steps below:

1. Data Cleaning:
   Run Data_Cleaning.py to clean the raw data (immoweb_data.csv) and save the cleaned dataset as immoweb_data_cleaned.csv.

2. Data Analysis:
   After cleaning the data, run Data_Analysis.py to analyze the cleaned data and generate insights.

3. Dashboard:
   Use dashboard.py to create a visualization dashboard by running:

   streamlit run dashboard.py

   This will start a local server and allow you to interact with the data via an interactive dashboard.

Deliverables

Step 1: Data Cleaning

The detailed explanation of the Data Cleaning process is in the Jupyter notebook real_estate_data_cleaning.ipynb, contributed by LAI Edoardo.

• Loaded and explored the dataset.
• Identified and visualized missing values.
• Identified critical columns to keep and drop, with explanations.
• Handled missing values by imputing data based on grouped analysis.
• Categorized variables to make them easier for analysis.
• Removed outliers and visualized their impact through histograms and boxplots.
• Prepared data for future modeling tasks.

Step 2: Data Analysis

The detailed explanation of Data Analysis is in the Jupyter notebook real_estate_data_analysis.ipynb, contributed by BOUBAKEUR Manel.

• Dataset Overview: The dataset has 16 columns and 16,631 rows.

• Correlation Analysis:

  • The strongest correlations with property price are:
    • Living Area (0.43)
    • Number of Rooms (0.33)
  • The weakest correlations with price are:
    • Lift (0.02)
    • Type of Property (0.02)
    • Swimming Pool (0.03)
    • Garden (0.04)

• Qualitative and Quantitative Variables:

  • We analyzed both types of variables and discussed the transformation of qualitative variables into numerical values for analysis.

Step 3: Data Interpretation

In this step, the analysis was communicated through tables and graphs by BOUBAKEUR Manel and FOMICHOV Andrii.

Key Findings:

1. Strong Positive Correlations:

   • Type_of_Property ↔ Subtype_of_Property (0.71)
   • Living_Area ↔ Number_of_Rooms (0.72)
   • Surface_area_plot_of_land ↔ Type_of_Property (0.73)

2. Moderate Positive Correlations:

   • Number_of_Rooms ↔ Type_of_Property (0.57)
   • Living_Area ↔ Type_of_Property (0.60)

3. Weak Positive Correlations:

   • Fully_Equipped_Kitchen ↔ State_of_the_Building (0.25)
   • Terrace ↔ Fully_Equipped_Kitchen (0.15)

4. Price vs Property Analysis:

   • The price distribution and mean values were plotted, identifying the most common price bins and the distribution across different types of properties.

   The price distribution shows a significant number of observations in the price range of 300,000 to 400,000 euros.

5. Price by Municipality:

   • The analysis of municipalities helped identify the most and least expensive areas in Belgium, Wallonia, and Flanders.

   Most Expensive Municipality (Belgium):

   • Knokke-Heist: Average Price = 601,451.55 EUR | Price per sqm = 7,464.10 EUR

   Least Expensive Municipality (Belgium):

   • Vaux-sur-Sûre: Average Price = 125,000.00 EUR | Price per sqm = 657.89 EUR

   The full breakdown for Wallonia and Flanders is provided in the project.

Graphs

Here are the visualizations related to the project:

• Correlation Heatmap for Apartment

• Correlation Heatmap for Houses

• Distribution of Property Prices

• Heatmap of Variable Correlations

• Heatmap of Correlation Between Variables and Price

• Number of Properties by Location

• Number of Properties by Province

• Outliers

• Percentage of Missing Values

• Price vs Area

• Price for Apartments

• Price for Houses

• Price for Properties

• Properties Price Distribution

Contributors

• BOUBAKEUR Manel – Data Analyst & Developer
• LAI Edoardo – Data Cleaning & Preprocessing
• FOMICHOV Andrii – Data Visualization & Interpretation