THULab/crypto_trading_bot_data

Crypto Trading Bot Data (TsFile)

Apache TsFile version of jponfiru/crypto-trading-bot-data.

Overview

OHLCV market data and derived features for a Binance spot trading bot: 1-minute klines for the top-20 USDT pairs (2024 onward), 5-minute klines for 2026, 1-hour klines, funding rates, open interest, macro indicators, the Fear & Greed index, BTC dominance, and a labelled feature table for the project's Model D.

• TsFile shards: 37 (1054.2 MB total, 29,706,465 rows across 9 table(s)).
  • btc_dominance — 6 rows in 1 shard(s), 2026-05-18 01:06:09.359000 → 2026-05-21 12:34:43.211000 (UTC)
  • fear_greed — 3,028 rows in 1 shard(s), 2018-02-01 00:00:00 → 2026-05-21 00:00:00 (UTC)
  • features_d — 219,043 rows in 1 shard(s), 2024-01-09 08:00:00 → 2026-05-10 14:00:00 (UTC)
  • funding_rates — 57,360 rows in 1 shard(s), 2024-01-01 00:00:00 → 2026-05-21 12:00:00 (UTC)
  • klines_1h — 383,190 rows in 1 shard(s), 2024-01-01 00:00:00 → 2026-05-11 15:00:00 (UTC)
  • macro_indicators — 54 rows in 1 shard(s), 2026-05-11 04:00:00 → 2026-05-21 07:00:00 (UTC)
  • ohlcv — 4,610,042 rows in 5 shard(s), 2024-01-01 00:00:00 → 2026-05-14 03:00:00 (UTC)
  • ohlcv_1m — 24,432,100 rows in 25 shard(s), 2024-01-01 00:00:00 → 2026-05-21 12:34:00 (UTC)
  • open_interest — 1,642 rows in 1 shard(s), 2026-05-18 02:00:00 → 2026-05-21 12:00:00 (UTC)

Conversion notes

• The source stores its series inside two SQLite databases; TsFile cannot read SQLite, so every table is streamed out of its database into a TsFile table of its own. Table list, row counts and skipped tables are recorded in conversion_summary.json.
• Tables with a pair (or symbol) column — ohlcv_1m, ohlcv, klines_1h, funding_rates, open_interest, macro_indicators and features_d — declare it as the TAG, so each instrument is one TsFile device with its own timeline; fear_greed and btc_dominance are single-device tables.
• Time = the source timestamp column at millisecond precision: ts_ms is already epoch milliseconds, ts is epoch seconds, and ohlcv.timestamp is an ISO-8601 string.
• No column and no row is dropped. ohlcv_1m (2024 onward) and ohlcv (2026) overlap; both are kept as separate tables rather than merged, and no duplicate key is removed because the two tables cover different timeframes.
• Not converted: download_state (data_2026.db), download_progress (data_extended.db) — bookkeeping tables recording the crawler's own progress, not time series.
• Not converted: features_extended.csv — 2.1 GB CSV of event features that was not downloaded for this conversion; the compact features_d.csv subset (219,043 rows) is included instead.
• The TsFiles are written with the zstd compressor. The Apache TsFile Java SDK reads every compressor, but the Python SDK (pip install tsfile) implements only UNCOMPRESSED, SNAPPY, GZIP and LZ4; opening these files with it returns no rows. Read them with the Java SDK, as shown below.

Usage

Read this dataset with the Java SDK. It is written with the zstd compressor. The Apache TsFile Java SDK reads every compressor, but the Python SDK (pip install tsfile) implements only UNCOMPRESSED, SNAPPY, GZIP and LZ4; opening this file with it yields no rows.

Read a converted file with the Apache TsFile Java SDK (org.apache.tsfile:tsfile):

import org.apache.tsfile.file.metadata.IDeviceID;
import org.apache.tsfile.read.TsFileReader;
import org.apache.tsfile.read.TsFileSequenceReader;
import org.apache.tsfile.read.common.Path;
import org.apache.tsfile.read.expression.QueryExpression;
import org.apache.tsfile.read.query.dataset.QueryDataSet;
import org.apache.tsfile.write.schema.MeasurementSchema;

import java.util.ArrayList;
import java.util.List;

String path = "btc_dominance.tsfile";
try (TsFileReader reader = new TsFileReader(new TsFileSequenceReader(path))) {
    for (String deviceId : reader.getAllDeviceIds()) {
        IDeviceID device = IDeviceID.Factory.DEFAULT_FACTORY.create(deviceId);
        List<Path> paths = new ArrayList<>();
        for (MeasurementSchema schema : reader.getMeasurement(device)) {
            String name = schema.getMeasurementName();
            if (name != null && !name.isEmpty() && !name.equalsIgnoreCase("Time")) {
                paths.add(new Path(device, name, true));
            }
        }
        QueryDataSet dataSet = reader.query(QueryExpression.create(paths, null));
        while (dataSet.hasNext()) {
            System.out.println(dataSet.next());
        }
    }
}

Source & license

• Original dataset: jponfiru/crypto-trading-bot-data
• License: mit (as declared by the original dataset).