Case DescriptionDoes information systems really reduce the daily workload of business people and improve productivity? How to transform from providing "passive" service to providing "active" service according to customer perception, truly realizing the effective management of information system service by Grid enterprises? How to construct a set of information system customer service perception model suitable for enterprise, through the model to accurately locate the problems existing in Information system customer service process, and through the establishment of information System customer service control system, continuously improve and optimize operation and maintenance services, improve customer service level, improve information system customer service satisfaction? has become an important work for enterprises to improve the level of informatization work effectively.
Case AnalysisInformation System customer service perception refers to the customer experience and feelings of information systems, reflecting the current quality of information systems and user expectations of the gap. In the comprehensive evaluation of customer service perception of information system, it involves a lot of complex phenomena and the interaction of many factors, moreover, there are a lot of fuzzy phenomena and fuzzy concepts in the evaluation. Therefore, in the comprehensive evaluation, some scholars use the method of fuzzy comprehensive evaluation to quantify, evaluate the information System customer service awareness level, and has achieved some results. However, using this method to model the customer service perception of information system, the determination of the weights of each input index requires expert knowledge and experience, which has great subjectivity and obvious defects, so it is not applicable. The use case is the subjective perception evaluation of the service effect of information system using the subject of information system, then the perceptual results of each influence factor are fused to the overall satisfaction rating of the information system, and the fuzzy input signal is fused. Fuzzy neural network combines the advantages of fuzzy evaluation method and neural network evaluation method, which has obvious advantages in solving such problems.
Predictive Modeling
operation Step One: Evaluation Index system DesignThe Information System Customer Service perception Evaluation Index is based on the following principles: 1 The Evaluation index can reflect the user's service evaluation of information system. 2) sample data is easy to collect, that is, the evaluation index data can be perceived by users. 3) through the improvement of these evaluation indicators, can really locate the problems existing in the process of information system services, to achieve continuous improvement and optimization of information systems. Based on the above principles and in-depth communication with business personnel, finally established the Information System Customer Service Perception Evaluation Index System, the index system covers the system itself and the system operation and maintenance of the indicators, a total of 6 first-class indicators and 18 two-level indicators, 16.
operation Step Two: sample preparationUser-perceived sample data mainly through the company's personal survey, a total of 19 business units, 5 categories of posts, 21 application systems, after the data preprocessing, the sample data of 19 business systems used for predictive modeling, and arbitrarily retain two business system data for model validation.
Action Step Three: attribute selectionAttribute selection, also known as attribute reduction, refers to the removal of irrelevant and redundant attributes based on the original value of the specific application data, and the selection of the most small subset of attributes to form subsets. This approach improves the quality of the data and speeds up learning, and attribute selection is an important part of the machine learning process. In the broad sense, the attribute selection algorithm can be divided into two algorithms, filter and Embedding mode (Wrapper).FCBF (Fast correlation-based Feature Select ion)belongs to the latter, so it has a certain advantage in processing the perceptual evaluation data with large attribute dimension. In general, if a feature is sufficiently correlated with a class, and its relevance to any other feature does not reach a certain level, it is considered a good feature for this class.FCBFTo use against uncertainty(symmetrical uncertainty,su)As a measure, the use ofSUValues for the property selection,SUValue in[0,1]Between1Indicates that two random variables can fully predict each other's values,0Indicates that two random variables are independent of each other.SUThe greater the value, the greater the superiority of its characteristics. Table 1 shows the attribute selection results for the FCBF search strategy based on the symmetry uncertainty evaluation sorting method. The sorting results of Table 1 also reflect the correlation between the evaluation indexes and the overall evaluation results. From table 1, it is indicated that the index which affects the overall evaluation of information system is the 7th, 16th, 17th and 8th attributes, which correspond to the stability of operation, the unblocked of complaint channel, the timeliness of fault handling and the timeliness of response. Attribute selection in perceptual evaluation modeling can not only find the minimum attribute set which is most suitable for customer satisfaction evaluation of information system, but also improve the performance of the algorithm. The experimental results show that the accuracy rate of the recognition is only slightly higher than the accuracy of the attribute set chosen by the attribute selection algorithm, but the latter is much more efficient in the algorithm. Therefore, attribute selection is a key step in the perceptual modeling process.
Operation Step Four: Model buildingConstruction process of fuzzy neural network Model 3:
operation Step Five: Model evaluationAfter the model training is completed, the perceptual evaluation model is validated by the production management system and the marketing management system respectively. Table 2 is the result of data fusion for the evaluation indicators of the two systems. Table -Results after the fusion of evaluation indicators
| Serial number |
Perceptual Evaluation Model |
Model Evaluation Results |
| Production Management System |
Marketing Management System |
| 1 |
Regression analysis |
3.7586 |
3.9299 |
| 2 |
BP Neural network |
3.9367 |
3.9644 |
| 3 |
RBF Neural Network |
3.8080 |
3.8104 |
| 4 |
FNN Neural Network |
3.5736 |
3.6386 |
In this paper, in addition to the fuzzy neural network to complete the information System Customer service perception evaluation modeling, and through the regression analysis, BP neural network, RBF Neural network modeling, the different model algorithm modeling results are compared and analyzed. The predictive evaluation results of different model algorithms are shown in table 3. Table theEvaluation results of different algorithms
| Serial number |
Perceptual Evaluation Model |
Model Evaluation Results |
| Production Management System |
Marketing Management System |
| 1 |
Regression analysis |
3.7586 |
3.9299 |
| 2 |
BP Neural network |
3.9367 |
3.9644 |
| 3 |
RBF Neural Network |
3.8080 |
3.8104 |
| 4 |
FNN Neural Network |
3.5736 |
3.6386 |
From table 3, the evaluation results of different perceptual models can basically reflect the user satisfaction evaluation of the application system, which algorithm is optimal, which can be measured by the RMS error of different algorithms in the model verification, and the results of the comparative analysis are shown in table 3.
From the modeling process and the results of verification, FNN Neural network is slightly slower than regression analysis, BP neural Network and RBF neural network algorithm, but in general, the prediction precision of FNN neural network is higher than regression analysis, BP neural Network and RBF neural network algorithm. This also embodies the advantage of fuzzy neural network for the modeling of customer service perception of information system.
On-Machine operating environment: www.tipdm.cn
FNN Fuzzy Neural Network--evaluation of information system customer service perception
