燃气多维度差异化用户画像数据

由于工商业用户和个人用户存在不同燃气使用习惯，即使同时个人用户每一户也一定存在不同的生活习惯、家庭结构、经济状况、用气习惯、设备条件等等差异，为了进一步分析用户的燃气使用习惯，为用户提供燃气安全提升等服务价值，通过数据算法，识别用户画像，得到分析结果。使用K均值聚类算法对每户的日用气量、月用气量进行聚类。分析用户的用气时间间隔信息，与聚类结果比较，标识长期不居住或正常居住的居住情况特征。将特征数字化处理，长期不居住标记为0.5，正常居住为1。
分析用户每日规律用气和月度总用气量情况，与聚类结果比较，标识青壮年用户，得到年龄层次特征。
分析签约情况、账户余额、缴费模式是否预付、逾期次数、总逾期金额，结合聚类结果评价是否属于多数，按100-100*归一化(ABS(逾期次数*(账户余额-逾期金额)))得到信用情况评分，归一化为计算所有欠费用户的逾期次数*金额，按从小到大排序，使用当前用户的逾期次数*逾期金额-最小值，再除以最大最小的差值。
根据月用气量/用气时当月天数与日用气量进行比较，如偏差在±20%以内，则用气趋势为平稳，小于20%或大于20%则用气趋势为下降、上升，对应的数值分别0.8,1.0,1.2。
按100-100*归一化（隐患数量*用气趋势*居住类型）得到用气安全评分。
按100-100*归一化(是否安装热水器*安装日期距今天数+是否安装壁挂炉*安装日期距今天数)得到终端设备情况评分。
户画像综合评分=每项得分/项数。

Given that commercial, industrial and residential users have distinct gas usage habits, and even individual residential households exhibit differences in living habits, family structures, economic conditions, gas usage patterns, equipment conditions and other aspects, to further analyze users' gas usage habits and deliver service values such as gas safety improvement, user portraits are identified through data algorithms to obtain analysis results.
The K-means clustering algorithm is used to cluster the daily gas consumption and monthly gas consumption of each household. The gas usage time interval information of users is analyzed and compared with the clustering results to identify the residence status characteristics of long-term non-residence or normal residence. The features are digitized, with long-term non-residence marked as 0.5 and normal residence marked as 1.
The regular daily gas usage and monthly total gas consumption of users are analyzed and compared with the clustering results to identify young and middle-aged users and obtain age group characteristics.
Analyze the contract status, account balance, whether the payment mode is prepaid, number of overdue payments, and total overdue amount, and evaluate whether the user belongs to the majority group combined with the clustering results. The credit score is calculated as 100 - 100 * normalization(ABS(number of overdue payments * (account balance - overdue amount))). The normalization is performed as: for all overdue users, sort their overdue payment count * overdue amount from smallest to largest, then use (current user's overdue payment count * overdue amount - minimum value) divided by (maximum value - minimum value).
Compare the monthly gas consumption divided by the number of days in the gas usage month with the daily gas consumption. If the deviation is within ±20%, the gas usage trend is stable; if the deviation is less than -20% or greater than +20%, the gas usage trend is decreasing or increasing, with corresponding values of 0.8, 1.0 and 1.2 respectively.
The gas safety score is calculated as 100 - 100 * normalization(number of hidden dangers * gas usage trend * residence type).
The terminal equipment status score is calculated as 100 - 100 * normalization( whether water heater is installed * number of days from installation date to current date + whether wall-hung boiler is installed * number of days from installation date to current date).
The comprehensive score of user portrait = sum of each item's score / number of items.

搜集汇总

数据集介绍

特点

燃气多维度差异化用户画像数据包含10万条记录，每周更新，通过算法分析用户用气习惯等多维度信息，形成用户画像评分，适用于燃气使用习惯分析和安全服务提升。