用户设备状态分析数据集合

核心采用“状态采集-行为画像-智能适配”三阶联动算法架构，细节如下：①状态采集算法：采用滑动窗口加权均值算法（工业设备5分钟窗口含60个数据点、家用设备1分钟窗口含120个数据点），按场景赋予数据权重，精准提取设备有效运行参数，避免场景干扰导致误判；②行为画像算法：以30天为周期，构建“使用频次-运行时长-故障概率”三维用户画像模型，通过K-means聚类算法划分用户类型（高频稳定型、低频间歇型、高风险操作型等）；③智能适配算法：结合国标要求及1000+用户场景实测数据，设置场景-用户双维度动态阈值（如高频工业用户：设备精度衰减≥2%FS触发预警；老年家用用户：告警未处置超30秒自动联动关阀+推送子女通知）。算法经250万条用户设备数据训练，第三方检测验证：状态识别准确率≥99.3%、用户画像匹配率≥97.5%、响应适配时间≤0.6秒，每月基于25万+条新增数据增量训练，适配用户行为与场景动态变化。

The core adopts a three-stage linkage algorithm architecture of 'state collection-behavior profiling-intelligent adaptation'. The details are as follows:
① State collection algorithm: Adopts the sliding window weighted average algorithm (for industrial equipment, the 5-minute window contains 60 data points, while for household equipment, the 1-minute window contains 120 data points). It assigns data weights according to scenarios to accurately extract effective operating parameters of equipment and avoid misjudgments caused by scene interference.
② Behavior profiling algorithm: Takes 30 days as a cycle to build a three-dimensional user profiling model of 'usage frequency-operating duration-failure probability', and classifies user types (high-frequency stable, low-frequency intermittent, high-risk operation, etc.) through the K-means clustering algorithm.
③ Intelligent adaptation algorithm: Combines national standard requirements and field test data from over 1000 user scenarios, and sets two-dimensional dynamic thresholds based on scenario and user dimensions. For example: for high-frequency industrial users, trigger an alert when equipment accuracy attenuation ≥2%FS; for elderly household users, automatically trigger valve closing and push notifications to their children if the alarm remains unaddressed for more than 30 seconds.
The algorithm is trained on 2.5 million data entries of user equipment, and verified by third-party testing: the state recognition accuracy is ≥99.3%, the user profiling matching rate is ≥97.5%, and the response adaptation time is ≤0.6 seconds. It conducts incremental training based on more than 250,000 new data entries every month to adapt to dynamic changes in user behaviors and scenarios.

搜集汇总

背景与挑战

背景概述

该数据集由贵州芯时代智能科技有限公司自行产生，聚焦燃气生产和供应业，数据规模为3G，每月更新。它通过三阶联动算法分析用户设备状态和行为，支持个性化运维、风险防控等服务，适用于设备用户和燃气运营企业，具有高准确率和动态适应性。