文章预览
简介 第一名方案主要由4个XGBoost模型（2个不同的标签）以及2个GRU模型（2个不同模型）组成。 验证策略 使用前500天数据进行训练并且将剩余的时间用做验证。 特征 county+datetime+data_block_id的组合特征。 fw_new_feature可以提升大致 0.1~0.2 # client_df client_df = client_df.drop([ "date" ]).sort([ "data_block_id" ], descending= False ) df = df.join(client_df.select([ "county" , "is_business" , "product_type" , "data_block_id" , "eic_count" , "installed_capacity" ]), how= "left" , on=[ "county" , "is_business" , "product_type" , "data_block_id" ]) # electricity electricity_df = electricity_df.drop([ "origin_date" ]).rename({ "forecast_date" : "datetime" }).with_columns([(pl.col( "datetime" ).str.to_datetime()+pl.duration(days= 1 )).alias( "datetime" ), (pl.col( "euros_per_mwh" ).abs()+ 0.1 ).alias( "euros_per_mwh" ),]) df = df.join(electricity_df[[ "data_block_id" , "datetime" , "euros_per_mwh" ]]
………………………………
