def update_a_u_map(reaction_list, a_u_map, insert_rate=0.001): """ 根据一段时间的交互记录,更新文章tag到用户tag之间的映射权重值 :param reaction_list: :param a_u_map: :param insert_rate: :return: """ for reaction in reaction_list: # process article tags reaction_a_id = reaction.reaction_a_id article = DAO_utils.get_article_by_id(reaction_a_id) article_tags = article.atags # get user tags reaction_user_id = reaction.reaction_user_id user = DAO_utils.get_user_by_id(reaction_user_id) utag_vec = user.user_tag_score_vec reaction_type = reaction.reaction_type reaction_date = reaction.reaction_date for atag_key in article_tags: # 采用稀释的方式改变映射权值 atag_value = article_tags[atag_key] # 文章本身的tag权值,用于衡量该篇文章的属性,0~1之间的值。 # 首先稀释原有的权值 a_u_inst = a_u_map[atag_key] for ukey in a_u_inst: a_u_inst[ukey] *= (1 - insert_rate) # 然后insert用户的tags for utag_key in utag_vec: if utag_key in a_u_inst: a_u_inst[utag_key] += utag_vec[utag_key] * insert_rate * atag_value else: a_u_inst[utag_key] = utag_vec[utag_key] * insert_rate * atag_value return a_u_map
def user_tagging(inst_user, reaction_list, reaction_type_weight, a_u_tagmap): """ 根据用户一段时间的交互记录,更新用户的u_tag分数 :param reaction_list:user_id等于当前用户id的一段时间内的交互记录 :param a_u_tagmap: :return: """ user_atag_vec = inst_user.user_atag_vec user_tag_score_vec = inst_user.user_tag_score_vec for reaction in reaction_list: weight = reaction_type_weight[reaction.reaction_type] a_id = reaction.reaction_a_id # todo a_map just for demo article = DAO_utils.get_article_by_id(a_id) for a_tag_key in article.a_tags: # 文章的tags应该是一个dict if a_tag_key in user_atag_vec: user_atag_vec[a_tag_key] += weight * article.a_tags[a_tag_key] else: user_atag_vec[a_tag_key] = weight * article.a_tags[a_tag_key] # 用户的atag_vec处理完毕,开始处理user_tag_score_vec # TODO 更好的权值赋值公式 for a_tag_key in user_atag_vec.keys(): for u_tag_key in a_u_tagmap[a_tag_key]: if u_tag_key in user_tag_score_vec: # TODO 是否需要每次都加? user_tag_score_vec[u_tag_key] = a_u_tagmap[a_tag_key][u_tag_key] * user_atag_vec[ a_tag_key] else: user_tag_score_vec[u_tag_key] = a_u_tagmap[a_tag_key][u_tag_key] * user_atag_vec[ a_tag_key] return user_tag_score_vec