Expense Prediction Of Complex Products And State Prediction Of Software Business Systems Based On Deep Learning

Forecasting is of great significance in real life,and the purpose of forecasting is to make things develop in a favorable direction through reasonable planning of future judgments.As a scientific and effective prediction method,deep learning is widely used in the field of prediction and has achieved good results.Therefore,based on the GAN in the deep learning model,this paper studies its application in the field of complex product cost estimation and software business system state prediction:(1)Traditional deep learning models require a large amount of data to train the neural network to ensure that the prediction results have good accuracy.Due to characteristics of complex products and the lack of complex product types,the cost forecast is in the case of a small sample,therefore,in this paper,a combination of GAN and CNN is used,and a GANCNN joint network model is proposed for complex product cost prediction in the case of small samples.This method solves the problem of small samples by generating new sample data from the generating network in the generative adversarial network,the discriminant network is used to extract the shallow features of the samples and share the shallow features to the prediction network composed of convolutional neural networks,finally,the prediction network and the discriminant network are used to constrain the generated network.The experimental results show that,compared with the traditional neural network-based complex product cost prediction results,the prediction results based on the GAN-CNN joint network model have smaller errors and higher accuracy.(2)Intelligent operation and maintenance is the mainstream direction of current operation and maintenance technology development.This paper proposes a GAN-SVRbased software business system state prediction method,which realizes intelligent operation and maintenance by predicting the possible state of the system in the future.According to the characteristic that the business system software information is time series data,this method first combines the GAN with the LSTM network,and uses the LSTM network as the generated network in the generative adversarial networks to simulate the distribution characteristics of real business system software information data,predict the future information of the business system software based on the generated samples;and then establish a mapping model of the business system software information and server resource consumption through SVR to obtain the future server resource consumption.(3)Based on the above two methods,the complex product cost forecasting system and the software business system status analyzing system are implemented respectively.The development architecture of the two systems and the design of the functional modules are introduced in detail.Finally,the functional interfaces of the two systems are shown by examples.