2019-09-02工作进展

上周五工作：
图片特征提取完成，mv-dssm网络训练中，当前网络训练效果参数：0.74, auc = 0.74, presition=0.81；

今天计划：
准备转正答辩

1. ugc图片特征

create table graph_embedding.hs_tmp_349 as select item_id, title, concat('https://img.alicdn.com/imgextra/',image_url) as url, image_position, type, status from tbcdm.dim_tb_itm_image where ds=max_pt('tbcdm.dim_tb_itm_image');

hs_dssm_dic_title_inf_10

drop table hs_tmp_350;
yes
create table hs_tmp_350 as select a.item_id as item_ida, b.item_id as item_idb, a.title, a.url, a.image_position, a.type, a.status from hs_tmp_349 a join hs_dssm_dic_title_inf_10 b on a.title = b.title;

create table hs_tmp_351 as select a.*, b.item_idb, b.title from hs_tmp_348 a join hs_tmp_350 b on a.item_id = b.item_ida;

insert overwrite table hs_tmp_353 select distinct item_idb, title, url from graph_embedding.hs_tmp_350 where image_position > 0 and image_position < 6 and type = 1 and status = 1;

create table graph_embedding.hs_tmp_354 as select a.item_idb, a.title, b.* from graph_embedding.hs_tmp_353 a join graph_embedding.hs_tmp_pic_1 b on a.url = b.url;

create table hs_tmp_355 as select a.*, b.url from hs_tmp_267 a join hs_tmp_353 b on a.video_id = b.item_idb;