Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,116 @@
|
|
1 |
-
---
|
2 |
-
license: mit
|
3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
---
|
4 |
+
|
5 |
+
## NIFTY-Feature-Enhanced
|
6 |
+
## Dataset Summary
|
7 |
+
|
8 |
+
NIFTY-Feature-Enhanced is a multi-modal, finance-focused dataset built on top of raeidsaqur/NIFTY
|
9 |
+
.
|
10 |
+
We enrich the original dataset with structured financial indicators, derived signals, temporal features, sentiment scores, embeddings, and event tags.
|
11 |
+
|
12 |
+
This makes it suitable for:
|
13 |
+
|
14 |
+
Predictive ML models (e.g., XGBoost, LSTMs, Transformers)
|
15 |
+
|
16 |
+
Financial NLP tasks (sentiment, RAG, semantic search)
|
17 |
+
|
18 |
+
Multi-modal research (numeric + textual features combined)
|
19 |
+
|
20 |
+
## Enrichments Added
|
21 |
+
## Temporal Features
|
22 |
+
|
23 |
+
day_of_week, month
|
24 |
+
|
25 |
+
## Market Indicators (parsed from context)
|
26 |
+
|
27 |
+
open, close, high, low, adj_close, volume, pct_change
|
28 |
+
|
29 |
+
Technical signals: macd, rsi, rsi_30, cci_30, dx_30, boll_ub, boll_lb, close_30_sma, close_60_sma
|
30 |
+
|
31 |
+
## Derived Financial Signals
|
32 |
+
|
33 |
+
daily_return = (close-open)/open
|
34 |
+
|
35 |
+
volatility = high-low
|
36 |
+
|
37 |
+
is_overbought (RSI>70), is_oversold (RSI<30)
|
38 |
+
|
39 |
+
## NLP Enrichments
|
40 |
+
|
41 |
+
news_embedding → 384-dim semantic vector (MiniLM)
|
42 |
+
|
43 |
+
finance_sentiment_scores (lexicon-based per-headline)
|
44 |
+
|
45 |
+
avg_finance_sentiment → aggregate sentiment per day
|
46 |
+
|
47 |
+
total_positive_hits, total_negative_hits
|
48 |
+
|
49 |
+
## Event Tags (regex-based)
|
50 |
+
|
51 |
+
mentions_policy, mentions_merger, mentions_earnings, mentions_commodity
|
52 |
+
|
53 |
+
## Rolling & Cross Features
|
54 |
+
|
55 |
+
rolling_close_3d, rolling_close_5d
|
56 |
+
|
57 |
+
rolling_volatility_5d
|
58 |
+
|
59 |
+
sma_crossover (30SMA vs. 60SMA)
|
60 |
+
|
61 |
+
sentiment_aligned_return = sentiment × pct_change
|
62 |
+
|
63 |
+
## Example Row
|
64 |
+
{
|
65 |
+
"date": "2010-01-26",
|
66 |
+
"open": 110.12,
|
67 |
+
"close": 109.77,
|
68 |
+
"volume": 147680200,
|
69 |
+
"macd": 0.8312,
|
70 |
+
"rsi_30": 59.84,
|
71 |
+
"daily_return": -0.0031,
|
72 |
+
"volatility": 1.12,
|
73 |
+
"is_overbought": 0,
|
74 |
+
"is_oversold": 0,
|
75 |
+
"avg_finance_sentiment": 0.007,
|
76 |
+
"mentions_policy": 1,
|
77 |
+
"mentions_merger": 0,
|
78 |
+
"mentions_earnings": 1,
|
79 |
+
"mentions_commodity": 1,
|
80 |
+
"rolling_close_3d": 110.95,
|
81 |
+
"rolling_close_5d": 112.31,
|
82 |
+
"sma_crossover": 1,
|
83 |
+
"sentiment_aligned_return": -2.1e-05,
|
84 |
+
"news_embedding": [0.036, -0.041, 0.082, ...] # 384-dim vector
|
85 |
+
}
|
86 |
+
|
87 |
+
## Use Cases
|
88 |
+
|
89 |
+
Financial prediction: Build ML models using enriched market + sentiment signals.
|
90 |
+
|
91 |
+
Financial NLP: Benchmark sentiment models, retrieval tasks, RAG pipelines.
|
92 |
+
|
93 |
+
Multi-modal ML: Combine embeddings + structured features for hybrid models.
|
94 |
+
|
95 |
+
Explainability studies: Investigate interactions between news tone and market moves.
|
96 |
+
|
97 |
+
## Citation
|
98 |
+
|
99 |
+
If you use this dataset, please also cite the original work:
|
100 |
+
|
101 |
+
NIFTY Dataset: Raeid Saqur, NIFTY: News-Informed Financial Trend Yield Dataset (2024)
|
102 |
+
|
103 |
+
@article{saqur2024nifty,
|
104 |
+
title={NIFTY: News-Informed Financial Trend Yield Dataset},
|
105 |
+
author={Saqur, Raeid and others},
|
106 |
+
journal={arXiv preprint arXiv:2405.09747},
|
107 |
+
year={2024}
|
108 |
+
}
|
109 |
+
|
110 |
+
## Acknowledgements
|
111 |
+
|
112 |
+
## Original dataset: raeidsaqur/NIFTY
|
113 |
+
|
114 |
+
## Enrichments by Naga Adithya Kaushik (GenAIDevTOProd)
|
115 |
+
|
116 |
+
## 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.
|