In view of the current problems of very few samples of electricity bill collection fraud and difficulty in identifying abnormal power payment users, this paper proposes a method to identify electricity bill collection...
详细信息
ISBN:
(纸本)9789819770038;9789819770045
In view of the current problems of very few samples of electricity bill collection fraud and difficulty in identifying abnormal power payment users, this paper proposes a method to identify electricity bill collection fraud based on nearest neighbor metric matching. First, by analyzing the characteristics of multi-source data on users' electricity consumption, the user payment behavior portrait categories are depicted. Subsequently, Gaussian process regression was used to build a user payment behavior prediction model based on normal user payment sample data. Finally, combined with dynamic multi-objective gravity search to obtain their nearest neighbor groups with different metrics, through abnormal probability distributionmatching, the abnormal payment thresholds and abnormal payment cycle thresholds of categories without abnormal payment user records are obtained, thereby generating electricity bill collection fraud screening rules and storing them on the blockchain as smart contracts, the accounts of abnormal transactions are checked through smart contracts. The experimental results show that the proposed method can promptly and accurately identify abnormal paying users, thereby effectively reducing fraud in electricity bill channels.
暂无评论