Does basic house information reflect house's description?

---

Joanna Broniarek, Vikranth Ale, Matteo Cavalletti

The aim of this project was to perform a clustering analysis of house announcements in Rome from Immobiliare.it.

Data Source

Data scrapping started from this link: https://www.immobiliare.it/vendita-case/roma/?criterio=rilevanza&pag=1

Steps

1. Creating two datasets (Information Dataset and Description Dataset) according to data scrapped from websites. Retrieving at least 10000 announcements.

2. Clustering the house announcements using K-mean++ method.In order to choose the best number of clusters per dataset the Elbow method was used.

3. Comparison among clusters. Do both datasets will lead to similar clusters?

• Finding similar clusters according to the Jaccard-Similarity.
• Returning the 3-most similar couples of clusters (information clusters vs description clusters)
• Creating Wordclouds of house descriptions. The represented words extracted from the description of the houses that are in the relative couple.

Jupyter Notebooks Descriptions

1. Main_HouseClustering - main notebook, where most of the analysis is contained
2. TFIDF_matrix - notebook for generating TFIDF matrix needed for Description Dataset.

Some functions used in analysis was located in the external files : collect_data.py, functions.py.

Folder DataFiles contains files, which were created and used during the analysis.

Technology

Python 3.6.4 (beautifulsoup, sklearn, wordcloud)