alpha_lab/data_ops_research/cta_1d

CTA 1D Parquet Dataset

This directory contains requirements for CTA (Commodity Trading Advisor) futures Parquet datasets used by alpha_lab.

Tables

cta_alpha158_1d

Alpha158 features for CTA futures.

• Source: dfs://daily_stock_run.stg_1day_tinysoft_cta_alpha159_0_7_beta
• Output: /data/parquet/dataset/cta_alpha158_1d/
• Columns: ~163 feature columns + code, m_nDate

cta_hffactor_1d

High-frequency factor features (8 columns).

• Source: dfs://daily_stock_run.stg_1day_tinysoft_cta_hffactor
• Output: /data/parquet/dataset/cta_hffactor_1d/
• Transformation: Pivot from long to wide format
  • Input columns: code, m_nDate, factor_name, value
  • Output columns: code, m_nDate, vol_1min, skew_1min, ... (8 features)
• Filter: Only include factor_name in [vol_1min, skew_1min, volp_1min, volp_ratio_1min, voln_ratio_1min, trend_strength_1min, pv_corr_1min, flowin_ratio_1min]

cta_dom_1d

Dominant contract mapping for continuous contracts.

• Source: dfs://daily_stock_run.dwm_1day_cta_dom
• Output: /data/parquet/dataset/cta_dom_1d/
• Filter: version = 'vp_csmax_roll2_cummax'
• Aggregation: GROUP BY m_nDate, code_init; SELECT first(code) as code

cta_labels_1d

Return labels for different return types.

• Source: dfs://daily_stock_run.stg_1day_tinysoft_cta_hfvalue
• Output: /data/parquet/dataset/cta_labels_1d/
• Filter: indicator in [twap_open1m@1_twap_close1m@1, twap_open1m@1_twap_open1m@2]
• Columns: code, m_nDate, indicator, value

Consumer

Used by: alpha_lab/cta_1d/src/loader_parquet.py

The alpha_lab project will create a parallel loader that reads from these Parquet tables instead of DolphinDB.