THULab/ohlcv_br_ativos

OHLCV Brazilian Stocks with Technical Indicators (TsFile)

Apache TsFile version of Wendelia/ohlcv-br-ativos.

Overview

Daily OHLCV series for five Brazilian stocks (PETR4, VALE3, ITUB4, WEGE3, MGLU3) together with the technical indicators the author computed for fine-tuning TimesFM 2.5 with LoRA: moving averages, RSI14, MACD, Bollinger bands, returns and volume ratios. Prices come from OpenBB (yfinance provider) and cover May 2022 – May 2026.

• Tables: itub4_daily, itub4_tecnicos, mglu3_daily, mglu3_tecnicos, petr4_daily, petr4_tecnicos, vale3_daily, vale3_tecnicos, wege3_daily, wege3_tecnicos
• Rows: 9,805 across 10 TsFile table(s).
  • itub4_daily: 1,080 rows, 2022-01-03 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • itub4_tecnicos: 881 rows, 2022-10-18 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • mglu3_daily: 1,080 rows, 2022-01-03 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • mglu3_tecnicos: 881 rows, 2022-10-18 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • petr4_daily: 1,080 rows, 2022-01-03 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • petr4_tecnicos: 881 rows, 2022-10-18 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • vale3_daily: 1,080 rows, 2022-01-03 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • vale3_tecnicos: 881 rows, 2022-10-18 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • wege3_daily: 1,080 rows, 2022-01-03 00:00:00 → 2026-04-30 00:00:00 (UTC)
  • wege3_tecnicos: 881 rows, 2022-10-18 00:00:00 → 2026-04-30 00:00:00 (UTC)

Schema (TsFile structure)

itub4_daily.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (7): open (double), high (double), low (double), close (double), volume (int64), split_ratio (double), dividend (double)

itub4_tecnicos.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (23): open (double), high (double), low (double), close (double), volume (int64), split_ratio (double), dividend (double), sma20 (double), sma50 (double), sma200 (double), rsi14 (double), macd (double), macd_sinal (double), macd_hist (double), bb_medio (double), bb_superior (double), bb_inferior (double), retorno (double), volatilidade_20d (double), volume_medio_20d (double), volume_rel (double), log_ret (double), range_pct (double)

mglu3_daily.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (7): open (double), high (double), low (double), close (double), volume (int64), split_ratio (double), dividend (double)

mglu3_tecnicos.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (23): open (double), high (double), low (double), close (double), volume (int64), split_ratio (double), dividend (double), sma20 (double), sma50 (double), sma200 (double), rsi14 (double), macd (double), macd_sinal (double), macd_hist (double), bb_medio (double), bb_superior (double), bb_inferior (double), retorno (double), volatilidade_20d (double), volume_medio_20d (double), volume_rel (double), log_ret (double), range_pct (double)

petr4_daily.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (6): open (double), high (double), low (double), close (double), volume (int64), dividend (double)

petr4_tecnicos.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (22): open (double), high (double), low (double), close (double), volume (int64), dividend (double), sma20 (double), sma50 (double), sma200 (double), rsi14 (double), macd (double), macd_sinal (double), macd_hist (double), bb_medio (double), bb_superior (double), bb_inferior (double), retorno (double), volatilidade_20d (double), volume_medio_20d (double), volume_rel (double), log_ret (double), range_pct (double)

vale3_daily.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (6): open (double), high (double), low (double), close (double), volume (int64), dividend (double)

vale3_tecnicos.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (22): open (double), high (double), low (double), close (double), volume (int64), dividend (double), sma20 (double), sma50 (double), sma200 (double), rsi14 (double), macd (double), macd_sinal (double), macd_hist (double), bb_medio (double), bb_superior (double), bb_inferior (double), retorno (double), volatilidade_20d (double), volume_medio_20d (double), volume_rel (double), log_ret (double), range_pct (double)

wege3_daily.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (6): open (double), high (double), low (double), close (double), volume (int64), dividend (double)

wege3_tecnicos.tsfile

• Time (INT64, milliseconds) — the source date timestamp as INT64 milliseconds.
• TAG: none — the table is a single device / single series.
• FIELD (22): open (double), high (double), low (double), close (double), volume (int64), dividend (double), sma20 (double), sma50 (double), sma200 (double), rsi14 (double), macd (double), macd_sinal (double), macd_hist (double), bb_medio (double), bb_superior (double), bb_inferior (double), retorno (double), volatilidade_20d (double), volume_medio_20d (double), volume_rel (double), log_ret (double), range_pct (double)

Conversion notes

• Each source CSV becomes its own TsFile table: <ticker>_daily (raw OHLCV) and <ticker>_tecnicos (OHLCV + technical indicators) for the five tickers, 10 tables in total. The per-ticker tables are the device dimension, so no TAG column is needed.
• Time is the date column (trading day) as INT64 milliseconds at 00:00 UTC.
• The *_daily tables cover a longer history than the *_tecnicos tables, whose rows start after the indicator warm-up period; both are kept, and no rows or columns are dropped.
• Not converted: dataset_lora_timesfm.parquet — consolidated copy of the five *_tecnicos.csv tables rounded to four decimals and missing 6 of their columns (sma200, bb_medio, retorno, volume_medio_20d, split_ratio, dividend); the CSVs are converted instead so no precision or column is lost.
• Not converted: resultados_backtest.csv, timesfm_lora_output/resultados_lora_teste.csv — back-test result tables (one row per ticker/horizon), not time series.

Usage

Install the Apache TsFile Python SDK (pip install tsfile) and read a converted file:

from pathlib import Path
from tsfile import TsFileReader

path = Path("itub4_daily.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    columns = [column.get_column_name() for column in table.get_columns()]
    print("columns:", columns)
    field_names = [
        column.get_column_name()
        for column in table.get_columns()
        if column.get_column_name() not in {"Time", "time"}
    ]
    if field_names:
        with reader.query_table(table_name, field_names[:3], batch_size=1024) as result:
            batch = result.read_arrow_batch()
            if batch is not None:
                print(batch.to_pandas().head())

Source & license

• Original dataset: Wendelia/ohlcv-br-ativos
• Source card title / publisher: OHLCV Brazilian Stocks with Technical Indicators
• License: mit (as declared by the original dataset).