Upload folder using huggingface_hub

README.md

@@ -0,0 +1,146 @@
+---
+license: cc-by-4.0
+task_categories:
+- tabular
+tags:
+- electoral
+- geography
+- brazil
+- duckdb
+- parquet
+- geocoding
+- mato-grosso
+pretty_name: Mato Grosso Electoral & Geolocated Voting Dataset (2014-2024)
+language:
+- pt
+- en
+size_categories:
+- 1M<n<10M
+configs:
+- config_name: votes_geo
+  data_files:
+  - split: train
+    path: data/votes_geo.parquet
+- config_name: votes
+  data_files:
+  - split: train
+    path: data/votes.parquet
+- config_name: candidates
+  data_files:
+  - split: train
+    path: data/candidates.parquet
+- config_name: polling_places
+  data_files:
+  - split: train
+    path: data/polling_places.parquet
+- config_name: section_polling_places
+  data_files:
+  - split: train
+    path: data/section_polling_places.parquet
+- config_name: dados_preditivos_mt_2026
+  data_files:
+  - split: train
+    path: data/dados_preditivos_mt_2026.parquet
+---
+
+# Mato Grosso Electoral & Geolocated Voting Dataset (2014-2024)
+
+## Dataset Summary
+This dataset contains the complete, cleaned, and geocoded electoral results of Mato Grosso (MT), Brazil, covering the election years **2014, 2018, 2020, 2022, and 2024**. It offers two levels of utilization:
+1. **Relational Schema**: Comprising normalized tables for votes per section, candidate registrations, physical polling places, and section-to-polling-place mappings.
+2. **Denormalized Enriched Table (`votes_geo`)**: A single unified file (~3.8M rows) that maps every vote record directly to its candidates' metadata and the precise physical location (with CEP, neighborhood, latitude, and longitude).
+
+This dataset is especially useful for spatial analysis of voting patterns, political geography, and research on electoral behaviors.
+
+---
+
+## Dataset Structure
+
+### 1. `votes_geo` (Unified Geolocated Votes)
+This is the fully denormalized dataset joining election results, candidates, and physical locations.
+* **Size**: 3,823,314 rows
+* **Key Columns**:
+  * `ano` (int64): Year of the election.
+  * `turno` (int64): Round (1 or 2).
+  * `uf` (string): Federation Unit ('MT').
+  * `municipio` (string): Name of the municipality.
+  * `zona` (int64): Electoral zone number.
+  * `secao` (int64): Voting section number.
+  * `cargo` (string): Office (e.g. PRESIDENTE, GOVERNADOR, DEPUTADO FEDERAL).
+  * `partido` (string): Candidate's political party acronym.
+  * `candidato` (string): Candidate name (or VOTO BRANCO / VOTO NULO).
+  * `votos` (int64): Quantity of votes in this section.
+  * `nome_local_votacao` (string): Polling place name (e.g., school name).
+  * `endereco_local_votacao` (string): Raw street address.
+  * `cep` (string): Resolved postal code.
+  * `bairro` (string): Resolved neighborhood.
+  * `latitude` (double): Geocoded latitude.
+  * `longitude` (double): Geocoded longitude.
+  * `geocode_status` (string): API result status (`success`, `zero_results`, etc.).
+  * `geocode_confidence` (string): Confidence level (`high`, `medium`, `low`).
+  * `has_manual_correction` (boolean): Flag indicating if coordinates were manually adjusted.
+
+### 2. `votes` (Raw Normalized Votes)
+Core table containing electoral results per section.
+* **Size**: 3,823,314 rows
+* **Fields**: Standard TSE columns normalized (e.g., `ano`, `turno`, `codigo_municipio_tse`, `zona`, `secao`, `codigo_cargo`, `cargo`, `numero_candidato`, `candidato`, `partido`, `votos`, `local_votacao`).
+
+### 3. `candidates` (Candidate Registry)
+Metadata about candidates running in Mato Grosso.
+* **Size**: 25,682 rows
+* **Fields**: `sq_candidato`, `ano`, `cargo`, `numero_candidato`, `candidato`, `partido`, `situacao_candidatura`, `situacao_turno`, etc.
+
+### 4. `polling_places` (Physical Locations)
+Deduplicated physical voting locations with geocoded coordinates.
+* **Size**: 2,142 rows
+* **Fields**: `polling_place_id`, `uf`, `codigo_municipio_tse`, `municipio`, `nome_local_votacao`, `endereco_local_votacao`, `endereco_normalizado`, `cep`, `bairro`, `latitude`, `longitude`, `geocode_confidence`, `geocode_status`, `google_place_id`, `google_formatted_address`.
+
+### 5. `section_polling_places` (Temporal Mapping)
+Resolves the mapping between a section (zona + secao) and a physical location (`polling_place_id`) for each election year.
+* **Size**: 36,643 rows
+* **Rationale**: TSE's internal `local_votacao` codes are not stable across years. This mapping tracks where each section was physically located in each year.
+
+### 6. `dados_preditivos_mt_2026` (2026 Predictions)
+Aggregated electoral patterns and prediction inputs.
+* **Size**: ~14,000 rows
+
+---
+
+## Data Origin & Ingestion Pipeline
+
+1. **Source Data**: Extracted from the official Open Data portal of the Brazilian Electoral Court (TSE) and IBGE's municipal meshes.
+2. **Deduplication Strategy**: To solve the instability of TSE's `local_votacao` code (where the same code can represent different schools across years, or different codes represent the same school), a unique `polling_place_id` was calculated using:
+   `SHA256(uf + codigo_municipio_tse + nome_local_votacao + endereco_normalizado)[:20]`
+3. **Geocoding**: Physical locations were processed using the Google Geocoding API.
+4. **Manual Auditing**: Low-confidence results or coordinates falling outside the Mato Grosso bounding box were manually verified and corrected. The final geolocation coverage is **99.86%**.
+
+---
+
+## Usage Instructions
+
+### Loading with Hugging Face Datasets (Python)
+To load the unified geolocated votes dataset directly in Python:
+```python
+from datasets import load_dataset
+import pandas as pd
+
+# Load the denormalized dataset (recommended for spatial queries)
+dataset = load_dataset("andrebadini/eleitoral-mt", name="votes_geo", split="train")
+df = dataset.to_pandas()
+print(df.head())
+```
+
+### Querying with DuckDB (SQL)
+DuckDB can read these Parquet files directly from the Hugging Face URL:
+```sql
+SELECT municipio, candidato, SUM(votos) AS total
+FROM read_parquet('https://huggingface.co/datasets/andrebadini/eleitoral-mt/resolve/main/data/votes_geo.parquet')
+WHERE ano = 2022 AND cargo = 'GOVERNADOR'
+GROUP BY municipio, candidato
+ORDER BY total DESC;
+```
+
+---
+
+## Licensing
+This dataset is distributed under the **Creative Commons Attribution 4.0 International (CC-BY-4.0)** license. You are free to share and adapt the data, provided that you give appropriate credit.

data/candidates.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b434db250c0887bcbf1fd019deaf2f014c7455e1161489d86f802908a8d2bbba
+size 1015597

data/dados_preditivos_mt_2026.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f609f50db75c06ab4b581db3b0c17701fb07b5076c063e3308ad46b101660885
+size 275223

data/polling_places.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a4bff3d070bacc6b128a721f2ab7b892189be455fc286bf3aa82f70df55bab0
+size 329193

data/section_polling_places.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e982c958c8db1c7e357ee970ec600d4261105412825e6fee6257efe46b438e8e
+size 705216

data/votes.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b8e0cd0d087eda1d60831f28b8e068dbf8691c87d18040b262da2fb63543564
+size 64375666

data/votes_geo.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c336c755d2e58c6df39f8ff96208b97b2f294f3cbf77b03d6c5b58805afd6658
+size 91736723