README.md

metadata

license: mit
dataset_info:
  features:
    - name: id
      dtype: string
    - name: context
      dtype: string
    - name: news
      dtype: string
    - name: conversations
      list:
        - name: role
          dtype: string
        - name: value
          dtype: string
    - name: label
      dtype: string
    - name: pct_change
      dtype: float64
    - name: day_of_week
      dtype: int64
    - name: month
      dtype: int64
    - name: open
      dtype: float64
    - name: close
      dtype: float64
    - name: rsi
      dtype: 'null'
    - name: macd
      dtype: float64
    - name: volume
      dtype: float64
    - name: high
      dtype: float64
    - name: low
      dtype: float64
    - name: adj_close
      dtype: float64
    - name: boll_ub
      dtype: float64
    - name: boll_lb
      dtype: float64
    - name: rsi_30
      dtype: float64
    - name: cci_30
      dtype: float64
    - name: dx_30
      dtype: float64
    - name: close_30_sma
      dtype: float64
    - name: close_60_sma
      dtype: float64
    - name: daily_return
      dtype: float64
    - name: volatility
      dtype: float64
    - name: is_overbought
      dtype: int64
    - name: is_oversold
      dtype: int64
    - name: date
      dtype: timestamp[us]
    - name: news_embedding
      list: float64
    - name: mentions_policy
      dtype: int64
    - name: mentions_merger
      dtype: int64
    - name: mentions_earnings
      dtype: int64
    - name: mentions_commodity
      dtype: int64
    - name: finance_sentiment_scores
      list: float64
    - name: avg_finance_sentiment
      dtype: float64
    - name: total_positive_hits
      dtype: int64
    - name: total_negative_hits
      dtype: int64
    - name: rolling_close_3d
      dtype: float64
    - name: rolling_close_5d
      dtype: float64
    - name: rolling_volatility_5d
      dtype: float64
    - name: sma_crossover
      dtype: int64
    - name: sentiment_aligned_return
      dtype: float64
  splits:
    - name: train
      num_bytes: 47360695
      num_examples: 1477
    - name: test
      num_bytes: 2527791
      num_examples: 317
    - name: valid
      num_bytes: 3603858
      num_examples: 317
  download_size: 31109624
  dataset_size: 53492344
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
      - split: valid
        path: data/valid-*

NIFTY-Feature-Enhanced

Dataset Summary

NIFTY-Feature-Enhanced is a multi-modal, finance-focused dataset built on top of raeidsaqur/NIFTY . We enrich the original dataset with structured financial indicators, derived signals, temporal features, sentiment scores, embeddings, and event tags.

This makes it suitable for:

Predictive ML models (e.g., XGBoost, LSTMs, Transformers)

Financial NLP tasks (sentiment, RAG, semantic search)

Multi-modal research (numeric + textual features combined)

Enrichments Added

Temporal Features

day_of_week, month

Market Indicators (parsed from context)

open, close, high, low, adj_close, volume, pct_change

Technical signals: macd, rsi, rsi_30, cci_30, dx_30, boll_ub, boll_lb, close_30_sma, close_60_sma

Derived Financial Signals

daily_return = (close-open)/open

volatility = high-low

is_overbought (RSI>70), is_oversold (RSI<30)

NLP Enrichments

news_embedding → 384-dim semantic vector (MiniLM)

finance_sentiment_scores (lexicon-based per-headline)

avg_finance_sentiment → aggregate sentiment per day

total_positive_hits, total_negative_hits

Event Tags (regex-based)

mentions_policy, mentions_merger, mentions_earnings, mentions_commodity

Rolling & Cross Features

rolling_close_3d, rolling_close_5d

rolling_volatility_5d

sma_crossover (30SMA vs. 60SMA)

sentiment_aligned_return = sentiment × pct_change

Example Row

{ "date": "2010-01-26", "open": 110.12, "close": 109.77, "volume": 147680200, "macd": 0.8312, "rsi_30": 59.84, "daily_return": -0.0031, "volatility": 1.12, "is_overbought": 0, "is_oversold": 0, "avg_finance_sentiment": 0.007, "mentions_policy": 1, "mentions_merger": 0, "mentions_earnings": 1, "mentions_commodity": 1, "rolling_close_3d": 110.95, "rolling_close_5d": 112.31, "sma_crossover": 1, "sentiment_aligned_return": -2.1e-05, "news_embedding": [0.036, -0.041, 0.082, ...] # 384-dim vector }

Use Cases

Financial prediction: Build ML models using enriched market + sentiment signals.

Financial NLP: Benchmark sentiment models, retrieval tasks, RAG pipelines.

Multi-modal ML: Combine embeddings + structured features for hybrid models.

Explainability studies: Investigate interactions between news tone and market moves.

Citation

If you use the NIFTY Financial dataset in your work, please consider citing our paper:

@article{raeidsaqur2024NiftyLM, title = {NIFTY-LM Financial News Headlines Dataset for LLMs}, author = {Raeid Saqur}, year = 2024, journal = {ArXiv}, url = {https://arxiv.org/abs/2024.5599314} }

Acknowledgements

Original dataset: raeidsaqur/NIFTY

Enrichments by Naga Adithya Kaushik (GenAIDevTOProd)

This makes NIFTY-Feature-Enhanced one of the most feature-rich financial datasets on Hugging Face, bridging numeric markets + NLP headlines for ML + GenAI research.