Bad data identification method based on density analysis coupled neural network prediction

The invention provides a bad data detection method based on density analysis coupled neural network prediction. The method mainly comprises the steps of obtaining original data, conducting dimensionless processing on the original data, calculating the distance between a sample and other samples, counting the density of surrounding points of the samples, and recording the samples with the density of the surrounding points smaller than a set value as bad samples. And learning, training and establishing a BP neural network model by using the rejected bad sample set data, carrying out loop test on the bad sample set through the neural network model, and finally obtaining a bad data set. Through coupling density analysis and a neural network model, bad data are preliminarily judged through the density of surrounding points of samples, then bad samples are predicted by using a neural network, and training samples and test samples are dynamically adjusted according to prediction results, so that the method is very s