Quick Description

Programmers : Zakaria Abdelmoiz DAHI, and Gabriel Luque.

About: This repositiory implements a machine learning pipeline devised in [1] for decision making in economics.

• [1] Dahi, Z.A., Luque, G., Alba, E. (2022). A Machine Learning-Based Approach for Economics-Tailored Applications: The Spanish Case Study. In: Jiménez Laredo, J.L., Hidalgo, J.I., Babaagba, K.O. (eds) Applications of Evolutionary Computation. EvoApplications 2022. Lecture Notes in Computer Science, vol 13224. Springer, Cham. https://doi.org/10.1007/978-3-031-02462-7_36

How 📗

- Depending on what component of the pipeline you want to execute, you have to navigate to the folder code and then to the folder dedicated to the component you want to execute.

Folders Hiearchy 📂

- Code:
    - Database: create and manipulate the database (Extraction module).
    - Profiling: Perform the profiling including the preprocessing steps.
    - Prediction: Performs the forecasting.
    - Animated_Maps: Produce an animated gif for a given metric.
    - Maps: Create a fixed map for a given metric.
    - old databases: previous databases before normalisation and pre-processing.
    - svg-map-aux: Other materials.
- Results:
    - Pre-profiling: contains the excel file of the pre-profiling.
    - Profiling: contains the excel files of the profiling for each of the data series.
    - Predictions: Contains the figure illustrating the forecasting.
    - Animated_Maps: Contains the animated maps of Spain for a given metric.

Demo 🎥

Tools:

• map_cli.py: generate a PNG file with a map with some data (python map_cli.py --help for all the options).
• map_gif.py: generate a GIF file with a map with some data (python map_gif.py --help for all the options).

Database structure (Original):

TL;DR: Given a place from locations (for example Málaga) and a data serie from series (for example male unemployment percentage), you can found the data about it in values using the id found in codes. Also codes contains information about when these data was obtained which allows to request new data to the INE DB.

• locations table: Places considered in our study (id and name of the place). Currently, we are considering only provinces.
• series table: Data series store in our database, for example, employment data (id, namesof the data serie, internal INE code for this serie and some description textabout it).
• codes table: Given a place from location and a serie from series, it stores the internal INE code to access this information, when it was updated (last_update for the date in INE DB and real_date for the date when our program gets those data). It also includes an id and a description text
• values table: Actual values for a pair <place (from locations), serie (frome series)>. These data are annotated with their year and period (the meaning of the period value can be a month, a semester, a quarter...).

Reduced Database structure:

FILE: db-reduced.db

A single table with clean data:

• Only series with data for all locations (33 series and 52 cities)
• Complete and consecutive years from 2003 to 2017 (15 years)
• 12 values per year (33 x 52 x 15 x 12 = 308 880 values)

Main Python Files

• database.py: This file contains the python declaration for the database (ORM system using SQLAlchemy). The method createDB create/open the SQLite database db.db returning an object to manipulate it.
• data.py: This is the main program to get the actual data from INE database.

Other Python Files

• test.py: Testing some operations over the database.
• testURL.py: Get all series in INE database
• newSerie.py: From a previous analysis (reduced one) generates the series.jsonto be downloaded.
• cleanData.py: Generate a clean version of the data (db-reduced.db)

Other Files:

• locations.sql: SQL sentences for creating and population localtions table.
• series.json: Data needed to query INE Database for the interested series.
• only-locs.db: SQLite Database only including the locations (for a clean restarting).
• db.db: Current SQLite Database.
• URL_analysis?.txt: All the data series including a city.
• URL_analysis?_reduced.txt: Series selected from URL_analysis?.txt (only includes the number of INE series and the name of the series removing the city name).

Map

Undocumented.

How to get a new series

To get the for some series:

• Search the series in URL_analysis.txt (this is the result of textURL.py).
• You have to include in series.json a new element with three values:
  • name: This is the name of the series (This is only information for the user)
  • code: This is the number which appears in the line Searching in X
  • text: This is the text removing the name of the location