{ "total_raw": 949, "n_csv_files": 33, "stages": [ { "stage": 0, "name": "原始指标", "count": 949, "rule": "33个CSV文件" }, { "stage": 1, "name": "数据可用性", "count": 136, "rule": "缺失率<30%, 覆盖>120月" }, { "stage": 2, "name": "单变量相关性", "count": 82, "rule": "|corr(feature, return)|>0.03" }, { "stage": 3, "name": "共线性去重", "count": 70, "rule": "组内pair |corr|<0.85" }, { "stage": 4, "name": "MI非线性筛选", "count": 70, "rule": "MI(feature; return)排序" }, { "stage": 5, "name": "最终选择", "count": 18, "rule": "经济学意义+因子组分配" } ], "final_features": [ { "feature": "Brent_spot", "factor_group": "Price", "mi_score": 0, "corr": 0 }, { "feature": "natgas_spot_henry", "factor_group": "Price", "mi_score": 0, "corr": 0 }, { "feature": "iron_ore_spot", "factor_group": "Price", "mi_score": 0, "corr": 0 }, { "feature": "rig_count_us_new", "factor_group": "Supply", "mi_score": 0.0041, "corr": -0.0327 }, { "feature": "supply_saudi", "factor_group": "Supply", "mi_score": 0, "corr": 0 }, { "feature": "us_oil_inventory_total", "factor_group": "Supply", "mi_score": 0, "corr": 0 }, { "feature": "pmi_us_mfg", "factor_group": "Demand", "mi_score": 0, "corr": 0 }, { "feature": "usd_index", "factor_group": "Demand", "mi_score": 0.0205, "corr": -0.0446 }, { "feature": "nonfarm_us", "factor_group": "Demand", "mi_score": 0, "corr": 0 }, { "feature": "ipi_us", "factor_group": "Demand", "mi_score": 0, "corr": 0 }, { "feature": "vix_lag1", "factor_group": "Risk_Geo", "mi_score": 0, "corr": 0 }, { "feature": "vix_lag2", "factor_group": "Risk_Geo", "mi_score": 0.0661, "corr": 0.1591 }, { "feature": "mom1m_lag1", "factor_group": "Technical", "mi_score": 0.0341, "corr": 0.0906 }, { "feature": "hist_vol_12m", "factor_group": "Technical", "mi_score": 0.065, "corr": 0.2184 }, { "feature": "rsi12m", "factor_group": "Technical", "mi_score": 0.0901, "corr": 0.2111 }, { "feature": "news_oil_sentiment", "factor_group": "Alternative", "mi_score": 0.0, "corr": 0.0311 }, { "feature": "news_geo_tone", "factor_group": "Alternative", "mi_score": 0.0566, "corr": -0.0511 }, { "feature": "news_article_volume", "factor_group": "Alternative", "mi_score": 0.017, "corr": 0.2619 } ], "rejected_examples": [ "mom1m", "bb_pct", "mom3m", "sota_ts_residual", "mom6m", "natgas_index", "cpi_japan", "hibor_3m", "gdp_australia_yoy", "bb_lower" ], "file_inventory": [ { "file": "commodities.csv", "n_cols": 31, "cols": [ "指标名称", "国际现货价格:黄金", "国际现货价格:白银", "国际现货价格:铜(A级)", "国际现货价格:铝", "国际现货价格:铅", "国际现货价格:锌", "国际现货价格:镍", "实际市场价:橡胶:全球:当月值", "实际市场价:可可豆:全球:当月值" ] }, { "file": "commodity_extended.csv", "n_cols": 38, "cols": [ "natgas_index", "agri_prices_橡胶", "agri_prices_可可豆", "agri_prices_罗布斯塔咖啡(豆)", "agri_prices_大豆", "agri_prices_大麦", "agri_prices_燕麦", "agri_prices_猪的肉", "agri_prices_家禽", "agri_prices_小麦" ] }, { "file": "financial_markets.csv", "n_cols": 22, "cols": [ "指标名称", "CBOE:波动率指数(VIX):收盘", "CBOE:波动性指数(VVIX):开盘", "标准普尔500波动率指数(VIX)", "指标名称.1", "美元兑欧元", "美元兑人民币", "美元兑日元", "美元兑瑞士法郎", "美元兑英镑" ] }, { "file": "geopolitical_shocks.csv", "n_cols": 27, "cols": [ "death_of_king", "pingjin", "crisis_of_iran", "turkey_to_iraq", "us_to_iran", "dep_of_dollar", "finance_crisis", "QE1", "debt_of_pigs", "QE2" ] }, { "file": "global_gdp.csv", "n_cols": 14, "cols": [ "gdp_china", "gdp_china_growth", "gdp_us", "gdp_japan", "gdp_germany", "gdp_uk", "gdp_france", "gdp_australia", "gdp_newzealand", "gdp_india_latam_国内生产总值[亿]" ] }, { "file": "iea_energy_policy.csv", "n_cols": 15, "cols": [ "title", "description", "status", "jurisdiction", "source", "datePromulgated", "yearEnded", "learnMore", "learnMoreLanguage", "dateModified" ] }, { "file": "live_market_data.csv", "n_cols": 16, "cols": [ "wti_futures_live", "brent_futures_live", "oil_vix_live", "vix_live", "uso_etf_live", "xle_etf_live", "xop_etf_live", "gold_futures_live", "natgas_futures_live", "heating_oil_live" ] }, { "file": "llm_event_scores.csv", "n_cols": 3, "cols": [ "event_name", "llm_intensity", "llm_sentiment" ] }, { "file": "macro_economics.csv", "n_cols": 41, "cols": [ "指标名称", "中国:制造业PMI", "美国:ISM:制造业PMI", "欧洲:制造业PMI", "中国:非制造业PMI:服务业", "美国:ISM:服务业PMI", "美国:工业生产指数:非季调", "欧洲:工业生产指数", "中国:工业生产指数(生产加权):季调", "美国:非农就业人数:季调:环比" ] }, { "file": "macro_extended.csv", "n_cols": 20, "cols": [ "cpi_china", "pmi_china", "ppi_china", "pmi_china_nonsvc", "cpi_us", "pmi_us", "ppi_us_同比", "ppi_us_同比.1", "pmi_us_svc", "employment_us_季调" ] }, { "file": "oil_composite_features.csv", "n_cols": 12, "cols": [ "指标名称", "us_oil_consumption", "india_oil_consumption", "china_oil_consumption", "china_oil_import", "china_refinery_yoy", "global_reserves", "murban_futures", "rig_count_us_new", "rig_count_opec_new" ] }, { "file": "oil_demand_by_product.csv", "n_cols": 42, "cols": [ "指标名称", "石油需求量:欧洲四国:液化石油气:当月值", "石油需求量:欧洲四国:汽油:当月值", "Unnamed: 3", "指标名称.1", "石油需求量:欧洲四国:石脑油:当月值", "Unnamed: 6", "指标名称.2", "石油需求量:欧洲四国:汽油:当月值.1", "石油需求量:欧洲四国:航空煤油:当月值" ] }, { "file": "oil_futures_positioning.csv", "n_cols": 39, "cols": [ "指标名称", "期货收盘价(连续):Murban原油:ICE", "ICE:原油:期货和期权:净持仓多头前8名占比", "ICE:原油:期货:掉期商:多头持仓:交易者数量", "ICE:原油:期货:净持仓多头前4名占比", "ICE:原油:期货和期权:掉期商:多头持仓", "ICE:原油:期货:管理基金:多头持仓:持仓数量占比", "ICE:原油:期货和期权:管理基金:套利持仓", "ICE:原油:期货:掉期商:多头持仓", "ICE:原油:期货和期权:管理基金:空头持仓:持仓数量占比" ] }, { "file": "oil_market_extended.csv", "n_cols": 28, "cols": [ "wti_futures_settlement", "brent_futures_settlement", "opec_basket", "micex_brent", "ice_brent", "oil_futures_close", "ice_positioning_1_ICE低硫轻质原油(WTI)期货期权持仓数量占比净持仓前8名多头", "ice_positioning_1_ICE低硫轻质原油(WTI)期货期权持仓数量占比净持仓前4名多头", "ice_positioning_1_ICE低硫轻质原油(WTI)期货期权持仓交易者数量报告头寸商业交易商多头", "ice_positioning_1_ICE低硫轻质原油(WTI)期货期权持仓数量占比非报告头寸空头总计" ] }, { "file": "oil_prices.csv", "n_cols": 23, "cols": [ "指标名称", "现货价:WTI原油", "Unnamed: 2", "Unnamed: 3", "指标名称.1", "现货价:Brent原油", "Unnamed: 6", "Unnamed: 7", "指标名称.2", "市场价(亚洲FOB):原油:阿曼" ] } ] }