For the intelligent dialogue system of customer service robots in power supply business halls, a large-scale dataset of power business user intents was constructed. The dataset includes 9 577 user queries and their labeling categories. First, the real voice data collected from the power supply business halls were cleaned, processed and filtered. In order to enable the data to drive the study of deep learning models related to intent classification, the data were labeled and augmented with high quality by the professionals according to the background knowledge of power business. In the labeling process, 35 types of service category labels were defined according to power business. In order to test the practicability and effectiveness of the proposed dataset, several classical models of intent classification were used for experiments, and the obtained intent classification models were put in the dialogue system. The classical Text classification model-Recurrent Convolutional Neural Network (Text-RCNN) was able to achieve 87.1% accuracy on this dataset. Experimental results show that the proposed dataset can effectively drive the research on power business related dialogue systems and improve user satisfaction.

With the rapid development and wide application of cloud storage system, many enterprise developers and individual users migrate their applications from traditional storage to public cloud storage system. Therefore, the performance of cloud storage system has become the focus of enterprise developers and individual users. The traditional test is difficult to simulate simultaneous access with enough users to the cloud storage system, complex to build and has a long test time with high cost. Besides, the evaluation results are unstable due to the network and other outside factors. In view of above critical problems, a kind of "cloud testing cloud" performance evaluation method was put forward for public cloud storage system. Public cloud storage system was evaluated by this method through applying a sufficient number of instances on the cloud computing platform. Firstly, a general performance evaluation framework was built with abilities such as dynamic instance application, automated deployment of assessment tools and load, controlling concurrent access to cloud storage system, automated instance release and evaluation results collection and feedback. Secondly, some multi-dimensional performance evaluation indicators were presented, covering different typical applications and different cloud storage interfaces. Finally, an extensible general performance evaluation model was put forward, which could evaluate the performance of typical applications, analyze the factors influencing cloud storage performance and be applied to any public cloud storage platform. In order to verify the feasibility, rationality, universality and expansibility of this method, these presented methods were applied to evaluate Amazon S3 cloud storage system, and then the accuracy of the evaluation results was verified by s3cmd. The results show that the evaluation output can provide reference comments for enterprise developers and individual users.

Focusing on high cost and complicated process of the fine particulate matter (PM _2.5) measurement system, a neural network model based on grey wolf optimizer algorithm was established. From the perspective of non-mechanism model, the daily PM _2.5 concentration in Shanghai was forecasted with meteorological factors and air pollutants, and the important factors were analyzed by mean impact value. To avoid the "over training" and ensure the generalization ability, the validation datasets were used to monitor the training process. The experimental results show that the most significant factors that affecting the PM _2.5 concentration are PM ₁₀, and then are the CO and the previous day's PM _2.5. Based on the datasets obtained from November 1, 2016 to November 12, the relative average error of the proposed model is 13.46%, the absolute average error is 8μg/m ³; the relative average error of it is decreased by about 3 percentage points, 5 percentage points and 1 percentage points compared with the prediction models based on Particle Swarm Optimization (PSO), BP neural network and Support Vector Regression (SVR). The neural network model based on the grey wolf optimizer algorithm is more suitable for forecasting PM _2.5concentration and air quality in Shanghai.

Since the efficiency of data forwarding in Vehicular Ad Hoc Network (VANET) is low, a data forwarding mechanism in VANET based on Software-Defined Network (SDN) was proposed. Firstly, a hierarchical architecture of SDN based VANET was designed. This architecture was consist of local controller and vehicular, it could implement the separation of control and data forwarding, and also could achieve high scalability, reliability and efficiency. Secondly, a new data forwarding mechanism was proposed, which used dynamic programming and binary search. Finally, compared with Ad Hoc On-demand Distance Vector routing (AODV), Destination Sequenced Distance Vector routing (DSDV), Dynamic Source Routing (DSR) and Optimized Link State Routing (OLSR) algorithm, the proposed algorithm could improve packet delivery fraction and end-to-end delay. Therein, the average increase of packet delivery fraction was about 100%, while the average reduction of end-to-end delay was about 20%. The simulation results show that the data forwarding mechanism in software-defined VANET can effectively improve the packet delivery and reduce the end-to-end delay.

Vehicular Ad Hoc Network (VANET) is a highly dynamic communication network, which means it's a great challenge to design a stable data dissemination mechanism. Applying Named Data Networking (NDN), which focused on the content of the data, to VANET could effectively relive the problems brought by the frequent change of network topology. Firstly, the message types and data structure of NDN were improved. Secondly, the way of establishing routes according to section was put forward with the combination of characteristics of VANET so as to reduce the cost of data dissemination. The simulation results show that compared to the traditional NDN algorithm which is applied to VANET data dissemination, Average Hit Rate (AHR) and Average Forward Times (AFT) can be significantly improved by VANET data dissemination mechanism based on NDN. Therein, the average increase of AHR is about 53 percent points, while the average reduction of AFT is about 0.4 times. Therefore, the improved VANET data dissemination mechanism can improve the efficiency of data dissemination by using the new routing method.

To solve the problem that the traditional query expansion used in professional domains suffers from the lack of semantic relations between expansion terms and original queries, a query expansion approach based on semantic vector representation was proposed. First, a semantic vector representation model was designed to learn the semantic vector representations of words from their contexts in corpus. Then, the similarities between words were computed with their semantic representations. Afterwards, the most similar words were selected from the corpus as the expansion terms to enrich the queries. Finally, a search system of biomedical literatures was built based on this expansion approach and compared with the traditional query expansion approaches based on Wikipedia or WordNet and the BioASQ participants along with the significant difference analysis. The comparison experimental results indicate that the proposed query expansion approach based on semantic vector representations outperforms the baselines, and the mean average precision increases by at least one percentage point; furthermore, the search system performs better than the BioASQ participants significantly.

The single feature based vehicle license plate localization algorithms are hard to be adapted to the complex environment. In order to solve the problem, a multi-feature fusion algorithm was proposed, which made use of multi-features such as edge, color and texture. The localization process was divided into two phases: Hypothesis Generation (HG) and Hypothesis Verification (HV). In HG, feature point detection algorithm and mathematical morphology were used as the primary techniques, and the character texture and color information of vehicle license plate were extracted as the features to generate the candidates. In HV, gray projection technology and constant feature of vehicle license plate were used to verify the candidates from HG, then the correct license plate was located. The experimental results show that the proposed algorithm can achieve the localization success ratio of 96.6% and the precision of 95.4% in the testing image set in real environment. Moreover, the rationality and validity of the multi-feature fusion algorithm are verified.

Mobile devices are mostly computational complexity sensitive because of the limited computing resource. The BandWidth Extension (BWE) algorithm in audio codec standard of China for mobile communication named AVS P10 was proposed to improve the mobile audio quality, but the computational complexity of the algorithm is too high to implement in mobile devices. The original BWE algorithm processes was analyzed, and the main reason of high computational complexity was identified to be the frequently usage of time-frequency transformation. Based on the analysis, a computational complexity optimization scheme was proposed, which include algorithm optimization and code optimization. The complexity of the algorithm was reduced by reducing the call number of Fast Fourier Transform (FFT). And the time consumption of the algorithm was reduced by some methods, such as sacrificing memory space for speed.The experimental results show that computation time consumption ratio of BWE module in encoder and decoder are decreased by 4.5 and 14.3 percentage points respectively, without reducing the overall audio codec subjective quality; the computational complexity of the algorithm is significantly reduced, which is beneficial to the application of the coding and decoding algorithm in the field of mobile audio.

To solve the challenge of accurate user group discovering, a user topic discovering algorithm based on trust chain, which was composed by three steps, i.e., topic space discovering, group core user discovering and topic group discovering, was proposed. Firstly, the related definitions of the proposed algorithm were given formally. Secondly, the topic space was discovered through the topic-correlation calculation method and a user interest calculation method for topic space was addressed. Further, the trust chain model, which was composed by atomic, serial, and parallel trust chains, and its trust computation method of topic space were presented. Finally, the detail algorithms of topic group discovering, including topic space discovering algorithm, core user discovering algorithm and topic group discovering algorithm, were proposed. The experimental results show that the average accuracy of the proposed algorithm is 4.1% and 11.3% higher than that of the traditional interest-based and edge density-based group discovering methods. The presented algorithm can improve the accuracy of user group organizing effectively, and it will have good application value for user identifying and classifying in social network.

A dynamic Retransmission Timeout (RTO) algorithm: DRTO (Dynamic RTO) for solving the TCP packets out-of-order caused by LTE network handover was proposed. The essence of DRTO was to use the TCP packet sequence number to distinguish the old and the new packets. Hence the multiplicative factor calculated in the past traditional RTO could be replaced by the difference of the serial number. The algorithm did not need to modify the handover mechanism, which can solve the packet out-of-order between the first part of packets (the packets transferred from source eNB to target eNB) which was transferred before handover process and the second part of the packets (the packets sent by server) which was transferred after handover process. Finally, the DRTO algorithm was compared with the traditional RTO algorithm on NS-2 simulation platform. The simulation results show the DRTO algorithm is better than the traditional RTO algorithm in terms of throughput, the number of retransmission packets and latency.

Through studying software network behavior, a new system model for analyzing the software network behavior based on dynamic binary analysis and message semantics analysis was proposed. The system consisted of dynamic binary analysis module, message semantics analysis module and network behavior analyzer. With the dynamic binary analysis, the Application Programming Interface (API) functions and system functions called by software could be obtained; by using the dynamic taint analysis, the message semantics could be extracted. The experimental results show that, combining the dynamic binary analysis and message semantics extraction can be used for analyzing the software network behavior.

A fuzzy recognition algorithm to detect computer virus approximately was present.It could overcome the shortage of normal virus scanner-which could not detect unknown virus.Based on this method,a virus detect network model was designed.This model can detect virus in the on-line system,and it can also detect known and unknown computer virus by analyzing the program behavior.

The paper discussed a model of customer classification based on Fuzzy ISODATA in CRM system. This model was put into practice by adopting the featuring parameters of RFM (Recency, Frequency, Monetary) and using the method of maximal matrix element to ascertain the number of classification. The system was realized with ASP and Microsoft SQL server, has been successfully implemented in a large Dairy corporation and provides the scientific basis to treat customers differently.

The intelligence and human-computer interaction of existing answering system are not good enough. To overcome the shortages, this paper presented a complete intelligent question answering System (IQAS) implementation. To enhance the veracity of matching between question from student and questions in database, the algorithm of question similarity was researched. By using auto segmenting algorithm, question similarity transformed the relativity of collections. A good study model was found to optimize segmenting weights from BP model, a supervised-machine learning task. Test results show that the algorithm can help IQAS improve veracity and intelligence and has practical value.