A dynamic workload description method and a dynamic workload matching algorithm were proposed to improve the accuracy of workload matching in automatic parameter tuning systems, such as OtterTune, with static workload description. First, a dynamic workload description method was proposed to describe the workload changes more accurately. Then, for the problems such as irregular sequences in workload matching and that the Euclidean distance algorithm is no longer applicable, a dynamic workload matching algorithm using data alignment was proposed based on the Dynamic Time Warping (DTW) algorithm. Finally, the proposed methods were applied to OtterTune to form a dynamic workload-based tuning tool D-OtterTune (Dynamic workload based OtterTune), and several experiments were conducted on it. Experimental results showed that, compared with the original method OtterTune, with the stable improvement in the workload matching accuracy for automatic parameter tuning brought by the dynamic workload matching method, D-OtterTune had the accuracy increased by 3%. It can be seen that D-OtterTune can have a significant impact on the overall business performance in data-intensive applications.

Concerning the problem that Stack Auto-encoder (SA) easily traps to overfitting, which may reduce the accuracy of spam classification, a modified SA method based on dynamic dropout was proposed. Firstly, the specificity of the spam classification was analyzed, and the dropout algorithm was employed in SA to handle overfitting. Then according to the fault of dropout algorithm that making some nodes be in the stall state for a long time, an improved algorithm of dropout was proposed. The static dropout rate was replaced by dynamic dropout rate which decreased with training steps using dynamic function. Finally, the dynamic dropout algorithm was used to improve the pretraining model of SA. The simulation results show that compared with Support Vector Machine (SVM) and Back Propagation (BP) neural network, the average accuracy of the modified SA is 97.66%. And the Matthews correlation coefficient of every dataset is higher than 89%. Matthews correlation coefficient of the modified SA on every dataset is 3.27%, 1.68%, 2.16%, 1.51%, 1.58% and 1.07% higher than that of the conventional SA separately. The experimental results show that the modified SA using dynamic dropout has higher accuracy and better robustness.

Specific to the need of performance prediction in communication networks, a connection rate prediction method based on fuzzy Auto-Regressive (AR) model was proposed and improved, and the fuzzy AR model based on adaptive fitting degree threshold was studied. The median filtering method was applied to pre-process the data of fuzzy AR model. On this basis, for the uncertain thresholds of some applications, the fitting degree threshold formula was added to the prediction model to make it adaptive. The simulation results show that the predistion method based on fuzzy AR model can be used to predict the connection rate with a higher fitting degree.