streamfuels 0.1.7

Data processing and analysis tools for fuel market research

StreamFuels

StreamFuels is a collection of updated fuel sales datasets for forecasting, classification, and pattern analysis, focusing on petroleum derivatives, natural gas, and biofuels market across different regions of Brazil.

monthly_sales_state(): Monthly fuel sales data by state from the ANP database

yearly_sales_state(): Yearly fuel sales data by state from ANP database

yearly_sales_city(): Yearly fuel sales data by city from ANP database

monthly_operations_state(): Monthly oil production, NGL production, natural gas production, reinjection, flaring and losses, self-consumption, and available natural gas. It provides a comprehensive view of petroleum and gas operations.

fuel_type_classification() Comprises 14,032 time series, each with a fixed length of 12 observations (i.e., one year of sales) and eight possible class labels.

Installation

pip install streamfuels

After that you can import using the target python environment:

from streamfuels.datasets import DatasetLoader
loader = DatasetLoader()
result, flag = loader.yearly_sales_state()

df, metadata = loader.read_tsf(path_tsf=result)

Yearly sales of petroleum derivatives in the states of Brazil.

result, flag = loader.yearly_sales_state()

Monthly sales of petroleum derivatives in the states of Brazil.

result, flag = loader.monthly_sales_state()

Monthly oil and gas operations in the states of Brazil.

result, flag = loader.monthly_operations_state()

Yearly sales of petroleum derivatives in the cities of Brazil.

result, flag = loader.yearly_sales_city()

Fuel Type Classification dataset

df = loader.fuel_type_classification()

Experimental results (April 2026)

• Classification (1NN, RF, Rocket, QUANT, HIVE-COTE v1)
  Download Classification results

• Forecasting (ARIMA, ETS, N-HiTS, N-BEATS)
  Download Forecasting results

Interactive Example Notebooks

You can explore practical use cases of the library directly via Google Colab or Binder.

---

Fuel Type Classification

• Demonstrates how to use classification algorithms to identify the fuel type.

---

Time Series Forecasting

• Shows how to perform time series forecasting using statistical and machine learning models.

---

Dataset Loading and Visualization

• Explains how to load datasets and visualize key information graphically.

---

Motif Discovery and Visualization

• Demonstrates how to identify and visualize repeating patterns (motifs) in time series data.

---

Notes

• Colab is recommended for faster and more stable execution of Python notebooks.
• Binder may take longer to build environments depending on dependencies.