P值計(jì)算好像不太對(duì)
老師你好，感謝分享！有個(gè)小問題：
ICF中最后P值計(jì)算，是指與J最相似的K個(gè)物品，和用戶U操作過的物品的交集，把他們的s求和。這個(gè)應(yīng)該要遍歷所有物品吧？
但是視頻代碼中，取用戶操作過的前3個(gè)物品，再取它最相似的K個(gè)物品，似乎不太符合
我重寫了下，老師你看有問題么，非常感謝！
def?cal_recom_result_2(sim_info,user_click):
????"""
????recom?by?item?collaboritive?filter
????Args:
????????sim_info:item?sim?dict
????????user_click:user?click?dict
????Return:
????????dict,key:userid?value?dict,?value_key?itemid,value_value?recome_score
????"""
????topk?=?5
????recom_info?=?{}
????for?user?in?user_click:
????????click_list?=?user_click[user]
????????recom_info.setdefault(user,{})
????????for?itemid_i,sim_item?in?sim_info.items():
????????????for?itemid_j,sim_score?in?sim_item[:topk]:
????????????????if?itemid_j?not?in?click_list:
????????????????????continue
????????????????recom_info[user].setdefault(itemid_j,0)
????????????????recom_info[user][itemid_j]?+=?sim_score
????return?recom_info
2018-12-11
您好，取了top幾個(gè)物品是由于有時(shí)效性，推薦過程中，我們不可能把用戶去年操作過的物品也拿過來做itemcf。因?yàn)殡S著時(shí)間推移?？赡茉缫呀?jīng)不喜歡之前的物品了。所以召回過程中考慮到這一點(diǎn)。謝謝。
2019-01-25
for?user_id,?item_list?in?user_click: ????recom_result.setdefault(user_id,?{}) ????for?itemi,?sim_dict?in?sim_info: ????????for?itemj,?val?in?sim_info[itemi][:k]: ????????????if?itemj?not?in?item_list[:recent_click_num]: ????????????????continue ????????????recom_result[user_id].setdefault(itemi,?0) ????????????recom_result[user_id][itemi]?=?val上面那個(gè)不對(duì)，應(yīng)該是這個(gè)
2019-01-25
recent_click_num?=?3 k?=?3 for?user_id,?item_set?in?user_click.items(): ????recom_result.setdefault(user_id,?{})???????? ????for?item_id1?in?item_set[:recent_click_num]: ????????if?item_id1?not?in?sim_info: ????????????continue ????????for?item_id2,?val?in?sim_info[item_id1]: ????????????if?item_id1?not?in?sim_info[item_id2][:k]: ????????????????continue ????????????recom_result[user_id].setdefault(item_id2,?0) ????????????recom_result[user_id][item_id2]?=?val return?recom_result2019-01-25