Welcome to the MARIA Project!
This repository contains all the information needed to reproduce our research, which is dedicated to extracting, transforming, and analyzing data on domestic violence in Brazil obtained from the Women’s Service Center (Call 180). The project also features interactive dashboards created with Tableau Public to visualize trends, patterns, and insights related to this critical issue.
MARIA (Mapping and Analysis of Risks and Insights on Abuse Against Women) is a comprehensive ETL (Extract, Transform, Load) pipeline designed to process, analyze, and enrich public data on domestic violence. The goal is to provide a clear understanding of the dynamics of abuse against women, supporting informed decision-making, awareness campaigns, and public policy development.
- Call 180: Open data provided by Brazil's Ministry of Human Rights and Citizenship.
- Period: 2014–2024 (excluding the 2019 balance sheet file containing consolidated summary data).
- Format:
.csvfiles categorized by year/semester.
- 2022 Census Data: Used for demographic analysis, including population size, education levels, race, and regional distribution.
The data processing pipeline, implemented in Python, is divided into three main stages:
-
Stage 1: Data Standardization
- Normalizing column names and formatting values (uppercase, no accents).
- Filtering records specific to domestic violence.
- Removing duplicates from data files post-2020.
-
Stage 2: Dataset Integration and Cleaning
- Merging datasets into four distinct groups based on changes in data collection formats:
- 2014 to November 2018.
- December 2018 to the end of 2019.
- First half of 2020.
- Second half of 2020 to 2023.
- Standardizing names of countries, cities, and professions.
- Merging datasets into four distinct groups based on changes in data collection formats:
-
Stage 3: Unified Dataset Creation
- Combining all cleaned datasets into a single, consistent dataset.
- Adding a "Categories" column to classify violations based on the Maria da Penha Law:
- Physical Violence
- Psychological Violence
- Sexual Violence
- Moral Violence
- Patrimonial Violence
Scripts also allow seamless updates whenever new data is released.
- Built with Tableau Public, the dashboards visualize trends in domestic violence cases over time and across regions.
- Access the dashboard here: MARIA Dashboard.
- Experimental clustering techniques:
- K-means and K-modes were applied to identify patterns and groupings within the data.
- This exploration aims to provide deeper insights into recurring trends and anomalies in the dataset.
- To reach a broader audience, data translation scripts (Python-based) convert the dataset and its outputs into English.
- This initiative supports global collaboration and awareness of domestic violence issues in Brazil.
MARIA/
├── Clusterização/ # Clustering analysis using K-means and K-modes.
├── Tradução/ # Scripts and outputs for translating data into English.
├── Tratamento/ # ETL pipeline scripts (Stages 1, 2, and 3 described above).
├── README.md # Project overview and details (this file).
Please refer to this work by citing the paper indicated below.
G. Zurawski, V. Pedroso, E. Patricio, G. Castro, A. F. Rollwagen, I. H. Manssour. Red Alert! Unveiling Insights in Violence Against Women through Interactive Data Visualization. Journal on Interactive Systems, Porto Alegre, RS, v. 16, n. 1, p. 516–531, 2025. https://doi.org/10.5753/jis.2025.5221.
We are members of the Data Visualization and Interaction Lab (DaVInt) at PUCRS:
- Gabriel Zurawski -- Undergraduate student in Computer Science.
- Vinícius Pedroso -- Undergraduate student in Data Science and Artificial Intelligence.
- Eduarda Patricio -- Undergraduate student in Computer Science.
- Giovanna Castro -- Undergraduate student in Computer Science.
- André F. Rollwagen -- PhD student in Computer Science.
- Isabel H. Manssour -- Researcher and Professor Coordinator of DaVInt.
More information can be found at https://www.inf.pucrs.br/davint/.