商户POI状态库数据

构建poi置信度模型，还未上线饿了么平台的潜在商户组成平台新的商机库，根据本数据，挖掘高活的POI数据供给商机库、进行商机库死店治理，清理被判定为死店的商户，帮助合理分配饿了么平台的营销、客户等资源供给。1、通过POI基础数据，采集POI最近一次拜访距今天数、POI最近一次人工核实为营业的距今天数、POI数据各来源零件的投票得分特征，根据零件的在线状态，在线为1，下线为-1，计算零件累加得分，2、将运单地址、wifi名称分别与POI中的门店名称、地址进行匹配，确认运单、wifi与poi的挂接关系，形成近7天、15天poi饿了么运单挂接数、最近一次运单挂接距今天数、poi最近有wifi挂接距今天数，3、基于以上这些静态特征构建前向神经网络，结合近12个月的poi运单挂接量：f_eleyd01_1-12，poi-wifi挂接量：f_wifi01_1-12，做特征归一化，构建transformer时序模型，时序模型输出与静态特征构建的前向神经网络进行特征拼接，输出预测结果，基于预测得分做置信度标签分层，输出字段new_confidence_level，预测模型得分>=0.45标签值为“高概活”，活店概率高，该POI可作为商机的供给，得分>=0.028且<0.45标签值为“中间”，活店概率为中间，得分<0.028标签值为“高概死”，该poi倒闭概率高，将poi下线并治理关联的有效商机。

We construct a POI confidence model. The potential merchants that have not yet launched on the Ele.me platform constitute the platform's new business opportunity repository. Based on this dataset, we mine high-activity POI data to supply the repository, perform dead store governance for the repository, clean up merchants judged as dead stores, and facilitate the rational allocation of marketing, customer and other resources of the Ele.me platform.
1. Collect features including the number of days since the last visit to the POI, the number of days since the last manual verification that the POI was operational, and the voting score characteristics of each source component of the POI data. Calculate the cumulative score of the components based on their online status: 1 for online, -1 for offline.
2. Match the waybill address and WiFi name with the store name and address in the POI respectively, confirm the association relationship between waybills, WiFi and POIs, and generate metrics including the number of Ele.me waybill associations for POIs in the past 7 and 15 days, the number of days since the POI's last waybill association, and the number of days since the POI's last WiFi association.
3. Build a feedforward neural network using the aforementioned static features. Combine this with the 12-month POI waybill association volume (f_eleyd01_1-12) and POI-WiFi association volume (f_wifi01_1-12), perform feature normalization, and construct a Transformer time-series model. Concatenate the output of the time-series model with the features extracted from the feedforward neural network to produce prediction results. Conduct confidence level labeling based on the prediction scores, with the output field designated as new_confidence_level. Specifically, when the prediction model score is ≥ 0.45, the label is "High Probability Active", signifying a high likelihood of the POI being an active store eligible to be supplied as a business opportunity. When 0.028 ≤ score < 0.45, the label is "Medium", indicating a moderate likelihood of the POI being an active store. When score < 0.028, the label is "High Probability Dead", signifying a high likelihood of the POI being closed, and the POI should be taken offline with its associated valid business opportunities subject to governance.

原始信息汇总

数据集概述

描述

浙江省数据知识产权登记平台是由浙江知识产权研究与服务中心推出的区块链数据知识产权登记系统。该系统支持数据知识产权登记、知识产权证书申请、原创作品登记确权、维权服务申请、维权证据出具、知识产权转让等多种场景，为创作者的知识产权提供全方位的保护。

关键词

• 区块链
• 知识产权
• 数据存证
• 知识产权存证
• 知识产权研究与服务中心
• 数据知识产权登记
• 浙江省数据知识产权登记平台

搜集汇总

数据集介绍

特点

该数据集为‘商户POI状态库数据’，包含1002条记录，每日更新，主要用于构建POI置信度模型，帮助饿了么平台挖掘高活跃度的潜在商户并进行死店治理。数据集包含多个字段，如POI ID、门店名称、地址等，并通过前向神经网络和transformer时序模型预测POI的活跃度，输出置信度标签。