feature-engineering

Name: feature-engineering
Author: agiprolabs

240installs53forks

How to install feature-engineering?

agiprolabs/claude-trading-skills/feature-engineering

$npx -y skills add agiprolabs/claude-trading-skills --skill feature-engineering

Installs into the current project.

Summary

Feature construction from market data for ML trading models including price, volume, on-chain, and microstructure features

Files · 1

View on GitHub

SKILL.md

1	# Feature Engineering for Trading ML
2
3	Feature engineering is the single highest-leverage activity in building ML trading
4	models. Model selection (XGBoost vs. neural net vs. logistic regression) matters far
5	less than the quality and diversity of input features. A simple model on great
6	features will outperform a complex model on raw prices every time.
7
8	This skill covers constructing, validating, and selecting features from market data
9	for use in classification (signal-classification) and regression models targeting
10	crypto/Solana token trading.
11
12	## Why Features Beat Models
13
14	Raw OHLCV data is non-stationary, noisy, and high-dimensional. Models trained
15	directly on price series will overfit. Feature engineering transforms raw data into
16	stationary, informative signals that capture distinct aspects of market behavior:
17
18	- Compression: Reduce thousands of price bars to dozens of descriptive statistics
19	- Stationarity: Convert non-stationary prices into stationary returns and ratios
20	- Domain knowledge: Encode trader intuition (support/resistance, volume climax)
21	as computable quantities
22	- Regime awareness: Features that behave differently in trending vs. ranging
23	markets help models adapt
24
25	## Feature Categories
26
27	### 1. Price Features
28
29	Derived purely from OHLCV price columns. These capture trend, momentum, and
30	volatility from the price series itself.
31
32	\| Feature \| Formula \| Lookback \|
33	\|---------\|---------\|----------\|
34	\| `log_return` \| `ln(close_t / close_{t-1})` \| 1 bar \|
35	\| `abs_return` \| `abs(log_return)` \| 1 bar \|
36	\| `return_volatility` \| `std(log_return, N)` \| 20 bars \|
37	\| `momentum_N` \| `close_t / close_{t-N} - 1` \| 5, 10, 20 \|
38	\| `acceleration` \| `momentum_5 - momentum_5[5]` \| 10 bars \|
39	\| `high_low_range` \| `(high - low) / close` \| 1 bar \|
40	\| `close_position` \| `(close - low) / (high - low)` \| 1 bar \|
41	\| `gap` \| `open_t / close_{t-1} - 1` \| 1 bar \|
42	\| `rolling_skew` \| `skew(log_return, N)` \| 20 bars \|
43	\| `rolling_kurtosis` \| `kurtosis(log_return, N)` \| 20 bars \|
44
45	### 2. Volume Features
46
47	Volume confirms or contradicts price movements. Divergences between price and
48	volume are among the most reliable signals in short-term trading.
49
50	\| Feature \| Formula \| Lookback \|
51	\|---------\|---------\|----------\|
52	\| `volume_ratio` \| `volume_t / mean(volume, N)` \| 20 bars \|
53	\| `volume_ma_ratio` \| `sma(volume, 5) / sma(volume, 20)` \| 20 bars \|
54	\| `obv_slope` \| `slope(OBV, N)` \| 10 bars \|
55	\| `vwap_deviation` \| `(close - VWAP) / VWAP` \| intraday \|
56	\| `volume_acceleration` \| `volume_ratio_t - volume_ratio_{t-1}` \| 21 bars \|
57	\| `buy_volume_ratio` \| `buy_volume / total_volume` \| 1 bar \|
58	\| `dollar_volume` \| `close * volume` \| 1 bar \|
59	\| `volume_cv` \| `std(volume, N) / mean(volume, N)` \| 20 bars \|
60
61	### 3. Technical Features
62
63	Standard technical indicators computed via `pandas-ta`. Use the `pandas-ta` skill
64	for full parameter documentation.
65
66	\| Feature \| Source \| Lookback \|
67	\|---------\|--------\|----------\|
68	\| `rsi` \| RSI(14) \| 14 bars \|
69	\| `macd_histogram` \| MACD(12,26,9) histogram \| 33 bars \|
70	\| `bb_position` \| `(close - BB_lower) / (BB_upper - BB_lower)` \| 20 bars \|
71	\| `bb_width` \| `(BB_upper - BB_lower) / BB_mid` \| 20 bars \|
72	\| `atr_ratio` \| `ATR(14) / close` \| 14 bars \|
73	\| `adx` \| ADX(14) \| 14 bars \|
74	\| `stoch_k` \| Stochastic %K(14,3) \| 14 bars \|
75	\| `cci` \| CCI(20) \| 20 bars \|
76	\| `mfi` \| MFI(14) \| 14 bars \|
77	\| `supertrend_direction` \| Supertrend direction (+1/-1) \| 10 bars \|
78
79	### 4. Microstructure Features
80
81	Derived from trade-level data (individual swaps/transactions). Require on-chain
82	or DEX API data.
83
84	\| Feature \| Description \|
85	\|---------\|-------------\|
86	\| `trade_count_ratio` \| Trades this bar / avg trades per bar \|
87	\| `avg_trade_size` \| Mean trade size in USD \|
88	\| `large_trade_pct` \| % of volume from trades > $10k \|
89	\| `unique_traders` \| Count of distinct wallet addresses \|
90	\| `buy_count_ratio` \| Buy trades / total trades \|
91	\| `trade_size_entropy` \| Shannon entropy of trade size distribution \|
92
93	### 5. On-Chain Features
94
95	Derived from blockchain state changes. Require Helius or Solana RPC data.
96
97	\| Feature \| Description \|
98	\|---------\|-------------\|
99	\| `holder_count_change` \| Change in unique holders over N periods \|
100	\| `whale_net_flow` \| Net tokens moved by top-10 holders \|
101	\| `token_velocity` \| Transfer volume / circulating supply \|
102	\| `liquidity_change` \| Change in DEX liquidity pool TVL \|
103
104	### 6. Cross-Asset Features
105
106	Capture relationships between the target token and broader market.
107
108	\| Feature \| Description \|
109	\|---------\|-------------\|
110	\| `sol_correlation` \| Rolling correlation with SOL price \|
111	\| `btc_beta` \| Rolling beta to BTC returns \|
112	\| `sector_momentum` \| Average return of tokens in same sector \|
113
114	### 7. Time Features
115
116	Cyclical encoding of calendar time. Use sin/cos encoding to preserve cyclical
117	continuity (hour 23 is close to hour 0).
118
119	```python
120	import numpy as np
121
122	hour_sin = np.sin(2 *

People who liked this also viewed

6 picks

Type

azure-costAzure cost management: query costs, forecast spending, optimize to reduce waste. WHEN: \"Azure costs\", \"Azure bill\", \"cost breakdown\", \"how much am I spending\", \"forecast spending\", \"optimizSkillsJul 2026275k1.3k
pricingWhen the user wants help with pricing decisions, packaging, or monetization strategy. Also use when the user mentions 'pricing,' 'pricing tiers,' 'freemium,' 'free trial,' 'packaging,' 'price increaseSkillsJul 202638k41k
stripe-integrationMulti-harness agentic plugin marketplace for Claude Code, Codex CLI, Cursor, OpenCode, GitHub Copilot, and Gemini CLISkillsJul 202611k38k
startup-financial-modelingMulti-harness agentic plugin marketplace for Claude Code, Codex CLI, Cursor, OpenCode, GitHub Copilot, and Gemini CLISkillsJul 20269.5k38k
tushare-finance专为 Claude Code 设计的 Tushare Pro 金融数据技能包，支持 220+ 个接口，用自然语言获取 A 股行情、财务数据、宏观经济等金融数据。SkillsJul 20265.0k75
azure-pricingFetches real-time Azure retail pricing using the Azure Retail Prices API (prices.azure.com) and estimates Copilot Studio agent credit consumption. Use when the user asks about the cost of any Azure seSkillsJul 20264.5k37k

feature-engineering

240installs53forks

View on GitHub

How to install feature-engineering?

agiprolabs/claude-trading-skills/feature-engineering

$npx -y skills add agiprolabs/claude-trading-skills --skill feature-engineering

Installs into the current project.

Summary

Feature construction from market data for ML trading models including price, volume, on-chain, and microstructure features

Files · 1

View on GitHub

SKILL.md

1	# Feature Engineering for Trading ML
2
3	Feature engineering is the single highest-leverage activity in building ML trading
4	models. Model selection (XGBoost vs. neural net vs. logistic regression) matters far
5	less than the quality and diversity of input features. A simple model on great
6	features will outperform a complex model on raw prices every time.
7
8	This skill covers constructing, validating, and selecting features from market data
9	for use in classification (signal-classification) and regression models targeting
10	crypto/Solana token trading.
11
12	## Why Features Beat Models
13
14	Raw OHLCV data is non-stationary, noisy, and high-dimensional. Models trained
15	directly on price series will overfit. Feature engineering transforms raw data into
16	stationary, informative signals that capture distinct aspects of market behavior:
17
18	- Compression: Reduce thousands of price bars to dozens of descriptive statistics
19	- Stationarity: Convert non-stationary prices into stationary returns and ratios
20	- Domain knowledge: Encode trader intuition (support/resistance, volume climax)
21	as computable quantities
22	- Regime awareness: Features that behave differently in trending vs. ranging
23	markets help models adapt
24
25	## Feature Categories
26
27	### 1. Price Features
28
29	Derived purely from OHLCV price columns. These capture trend, momentum, and
30	volatility from the price series itself.
31
32	\| Feature \| Formula \| Lookback \|
33	\|---------\|---------\|----------\|
34	\| `log_return` \| `ln(close_t / close_{t-1})` \| 1 bar \|
35	\| `abs_return` \| `abs(log_return)` \| 1 bar \|
36	\| `return_volatility` \| `std(log_return, N)` \| 20 bars \|
37	\| `momentum_N` \| `close_t / close_{t-N} - 1` \| 5, 10, 20 \|
38	\| `acceleration` \| `momentum_5 - momentum_5[5]` \| 10 bars \|
39	\| `high_low_range` \| `(high - low) / close` \| 1 bar \|
40	\| `close_position` \| `(close - low) / (high - low)` \| 1 bar \|
41	\| `gap` \| `open_t / close_{t-1} - 1` \| 1 bar \|
42	\| `rolling_skew` \| `skew(log_return, N)` \| 20 bars \|
43	\| `rolling_kurtosis` \| `kurtosis(log_return, N)` \| 20 bars \|
44
45	### 2. Volume Features
46
47	Volume confirms or contradicts price movements. Divergences between price and
48	volume are among the most reliable signals in short-term trading.
49
50	\| Feature \| Formula \| Lookback \|
51	\|---------\|---------\|----------\|
52	\| `volume_ratio` \| `volume_t / mean(volume, N)` \| 20 bars \|
53	\| `volume_ma_ratio` \| `sma(volume, 5) / sma(volume, 20)` \| 20 bars \|
54	\| `obv_slope` \| `slope(OBV, N)` \| 10 bars \|
55	\| `vwap_deviation` \| `(close - VWAP) / VWAP` \| intraday \|
56	\| `volume_acceleration` \| `volume_ratio_t - volume_ratio_{t-1}` \| 21 bars \|
57	\| `buy_volume_ratio` \| `buy_volume / total_volume` \| 1 bar \|
58	\| `dollar_volume` \| `close * volume` \| 1 bar \|
59	\| `volume_cv` \| `std(volume, N) / mean(volume, N)` \| 20 bars \|
60
61	### 3. Technical Features
62
63	Standard technical indicators computed via `pandas-ta`. Use the `pandas-ta` skill
64	for full parameter documentation.
65
66	\| Feature \| Source \| Lookback \|
67	\|---------\|--------\|----------\|
68	\| `rsi` \| RSI(14) \| 14 bars \|
69	\| `macd_histogram` \| MACD(12,26,9) histogram \| 33 bars \|
70	\| `bb_position` \| `(close - BB_lower) / (BB_upper - BB_lower)` \| 20 bars \|
71	\| `bb_width` \| `(BB_upper - BB_lower) / BB_mid` \| 20 bars \|
72	\| `atr_ratio` \| `ATR(14) / close` \| 14 bars \|
73	\| `adx` \| ADX(14) \| 14 bars \|
74	\| `stoch_k` \| Stochastic %K(14,3) \| 14 bars \|
75	\| `cci` \| CCI(20) \| 20 bars \|
76	\| `mfi` \| MFI(14) \| 14 bars \|
77	\| `supertrend_direction` \| Supertrend direction (+1/-1) \| 10 bars \|
78
79	### 4. Microstructure Features
80
81	Derived from trade-level data (individual swaps/transactions). Require on-chain
82	or DEX API data.
83
84	\| Feature \| Description \|
85	\|---------\|-------------\|
86	\| `trade_count_ratio` \| Trades this bar / avg trades per bar \|
87	\| `avg_trade_size` \| Mean trade size in USD \|
88	\| `large_trade_pct` \| % of volume from trades > $10k \|
89	\| `unique_traders` \| Count of distinct wallet addresses \|
90	\| `buy_count_ratio` \| Buy trades / total trades \|
91	\| `trade_size_entropy` \| Shannon entropy of trade size distribution \|
92
93	### 5. On-Chain Features
94
95	Derived from blockchain state changes. Require Helius or Solana RPC data.
96
97	\| Feature \| Description \|
98	\|---------\|-------------\|
99	\| `holder_count_change` \| Change in unique holders over N periods \|
100	\| `whale_net_flow` \| Net tokens moved by top-10 holders \|
101	\| `token_velocity` \| Transfer volume / circulating supply \|
102	\| `liquidity_change` \| Change in DEX liquidity pool TVL \|
103
104	### 6. Cross-Asset Features
105
106	Capture relationships between the target token and broader market.
107
108	\| Feature \| Description \|
109	\|---------\|-------------\|
110	\| `sol_correlation` \| Rolling correlation with SOL price \|
111	\| `btc_beta` \| Rolling beta to BTC returns \|
112	\| `sector_momentum` \| Average return of tokens in same sector \|
113
114	### 7. Time Features
115
116	Cyclical encoding of calendar time. Use sin/cos encoding to preserve cyclical
117	continuity (hour 23 is close to hour 0).
118
119	```python
120	import numpy as np
121
122	hour_sin = np.sin(2 *

People who liked this also viewed

6 picks

Type