In this project, I mined data of each of the 73 neighborhoods of Barcelona, including their demographics, the services they have (hospitals, pharmacies) as well as other variables like: extension of green areas, number of hotels, etc., in order to see similitudes or differences between neighborhoods. I also used these indicators to see if I was able to predict the income of these neighborhoods using different models (supervised and unsupervesied), like K-Means, DBSCAN, SVD or Decision Trees.

The project is organized in two self-explanatory jupyter notebooks (with html version). Please go through them, they will guide you throught the project.

• Part 1: Contains Aims, and ETL, and primary data analysis including PCA. (Notebook / Html)

• Part 2: Contains assesment of different models and global conclusions (Notebook / Html)

Example: Prediction of neighborhood income based on all available features, using the best Single Vector Decomposition (SVD) via GridSearch.

Income Legend:

Very High High Medium Low

Ground Truth

SVC Prediction

• SVC Test Accuracy: 0.45
• SVC Test Precision: 0.57
• SVC Test F1_score: 0.48

Check the html files for interactive maps!