Data verification method based on multi-source transfer learning

Abstract

The invention proposes a data verification method based on multi-source transfer learning. The method comprises the steps: extracting the quantity of site services corresponding to a source data set and a target training set and normalizing the data; constructing a weight-based SVR model through a transfer learning model SVR model and a radial basis function; initializing the source data and a site weight of a target province, performing the normalization, merging a normalized source data set, a normalized target training data set, a normalized business data quantity training set and a normalized service quantity, and obtaining a merged training set; establishing a prediction model through the merged training set and a normalized vector, and calculating model error parameters; carrying outthe iteration operation for many times, and calculating a final prediction model; obtaining the service quantity of a predicted site of the target province through a final prediction model, and denormalizing the service quantity of the predicted site. Compared with the prior art, the method improves the data quality, and saves the data resources.

Description

technical field [0001] The invention belongs to the category of transfer learning, and in particular relates to a data verification method based on multi-source transfer learning. Background technique [0002] State Grid Telecommunication Management System (TMS), as the second physical network of the power company, carries the core business of power grid operation and management, and is an important guarantee for the safety, stability and economic operation of the power grid. As the core management system of the power company's communication specialty, the TMS system has played a huge role in resource management, real-time monitoring, and operation management, and has also accumulated a large amount of data. The TMS system is stored in the form of a database, and each unit independently deploys a database server. It mainly includes TMS resource data, alarm data, work order data, and business data generated by internal modules; State Grid Communication Company and its branch...

AI Technical Summary

Problems solved by technology

These two multi-source migration learning algorithms will select the data source with the strongest correlation with the target data in each iteration, so that although the source data for migration can be guaranteed to be the most relevant to the target, they do not use information from other data sources, and in actual production The cost of each data source in is very high, and this operation wastes a lot of resources of the company

The problem of data quality in the TMS system has seriously affected the company's judgment and operation of the actual business, and the differences in data distribution and data volume in various regions have also brought challenges to the discovery of data quality problems

Images