kawsersikder/bangladesh-stock-market-dataset

Bangladesh Stock Market Dataset: 27 Years of Open-Source Dhaka Stock Exchange Data with Technical Indicators and Deep Learning Benchmarks

Author: Kawser Sikder

Overview

A comprehensive, open-source financial dataset covering 441 publicly traded instruments across 23 industry sectors of the Dhaka Stock Exchange (DSE), Bangladesh's principal securities market.

Dataset Variants

1. Unprocessed Data (Raw OHLCV)

Clean daily trading data with 6 columns: Date, Open, High, Low, Close, Volume.

2. Processed Data (With Technical Indicators)

Same data augmented with 14 important technical indicators:

• —Trend: SMA (10, 20, 50), EMA (12, 26)
• —Momentum: MACD, MACD Signal, MACD Histogram, RSI (14-day)
• —Volatility: Bollinger Bands (Upper, Middle, Lower), ATR (14-day)
• —Volume: On-Balance Volume (OBV)

Folder Structure

Dhaka Stock Exchange Dataset/
├── Unprocessed Data/                      # Raw OHLCV (441 CSVs, 23 sector folders)
├── Processed Data (With Indicators)/      # With 14 technical indicators (441 CSVs)
├── feature_engineering.py                 # Reproducible indicator computation script
├── colab_stock_predictor.ipynb            # Benchmark ML/DL pipeline (Google Colab)
├── technical_report.md                    # Full dataset description paper
└── README.md                             # This file

Sector Distribution

Benchmark: ML/DL Prediction Pipeline

A Google Colab notebook (colab_stock_predictor.ipynb) is included. It implements a two-phase pipeline on Google Colab's free T4 GPU:

Phase 1 — GPU Ensemble Screening (Top 70): 4-model ensemble using XGBoost, LightGBM, CatBoost, and a PyTorch MLP Neural Network.

Phase 2 — Time-Series Transformer (Top 15): Transformer Encoder with Positional Encoding, Huber Loss, and automated hallucination filters.

Preliminary Results: Phase 1 achieved 58-70% direction prediction accuracy. Phase 2 produced realistic 2-week return forecasts (+0.78% to +8.37%).

Quick Start

import pandas as pd

# Load a single stock
df = pd.read_csv("Processed Data (With Indicators)/Bank/ABBANK.csv")
print(df.head())
print(df.columns.tolist())

Use Cases

• —Time-series forecasting (LSTM, Transformer, etc.)
• —Technical analysis strategy backtesting
• —Sector rotation and momentum studies
• —Portfolio optimization in emerging markets
• —Transfer learning for other South Asian markets

Limitations

• —Survivorship bias: Only actively listed instruments (as of June 2026) are included.
• —No fundamental data: No earnings, dividends, or book value.
• —No intraday data: Daily OHLCV only.
• —Corporate actions: Stock splits and rights issues may not be fully adjusted.

Citation

If you use this dataset in your research, please cite:

@dataset{sikder2026bdstock,
  author    = {Sikder, Kawser},
  title     = {Bangladesh Stock Market Dataset: 27 Years of Open-Source Dhaka Stock Exchange Data with Technical Indicators and Deep Learning Benchmarks},
  year      = {2026},
  publisher = {Hugging Face},
  note      = {1.5M+ daily trading records, 441 instruments, 23 sectors, 1999-2026}
}

License

This dataset is released under the Creative Commons Attribution 4.0 (CC BY 4.0) license.