Data filtering algorithm based on attribute reduction and gene expression programming

Active distribution network as an important part of the construction of smart grid, the traditional means of protection are difficult to meet the active distribution network under the multi-source mass data transmission security and efficiency. In order to solve this problem, this paper proposes a data filtering algorithm based on attribute reduction and gene expression programming (DF-ARGEP), and improves the two algorithms. The attribute reduction (AR) is optimized on the basis of the degree of dependency, and the gene expression programming algorithm(GEP) has optimized the selection strategy. AR can drastically reduce the data complexity of active distribution networks, while GEP has the advantage of simple coding to solve complex problems. Thus the combination of them can be used to proactively and efficiently protect the data in an active distribution network. According to the experiment with multiple UCI data sets, the attribute reduction algorithm (AR-DO) based on dependency ranking is more efficient. The experiments on the same meteorological data show that the average time-consumption of GEP is shorter while its accuracy is higher, besides, comparing with traditional genetic algorithm, DF-ARGEP performs better.