Long-Term Natural Gas Consumption Forecasting Based on Analog Method and Fuzzy Decision Tree

Classic forecasting methods of natural gas consumption extrapolate trends from the past to subsequent periods of time. The paper presents a different approach that uses analogues to create long-term forecasts of the annual natural gas consumption. The energy intensity (energy consumption per dollar of Gross Domestic Product—GDP) and gas share in energy mix in some countries, usually more developed, are the starting point for forecasts of other countries in the later period. The novelty of the approach arises in the use of cluster analysis to create similar groups of countries and periods based on two indicators: energy intensity of GDP and share of natural gas consumption in the energy mix, and then the use of fuzzy decision trees for classifying countries in different years into clusters based on several other economic indicators. The final long-term forecasts are obtained with the use of fuzzy decision trees by combining the forecasts for different fuzzy sets made by the method of relative chain increments. The forecast accuracy of our method is higher than that of other benchmark methods. The proposed method may be an excellent tool for forecasting long-term territorial natural gas consumption for any administrative unit.

Dynamic programming based fuzzy partition in fuzzy decision tree induction

Fuzzy decision trees are one of the most popular extensions of decision trees for symbolic knowledge acquisition by fuzzy representation. Among the majority of fuzzy decision trees learning methods, the number of fuzzy partitions is given in advance, that is, there are the same amount of fuzzy items utilized in each condition attribute. In this study, a dynamic programming-based partition criterion for fuzzy items is designed in the framework of fuzzy decision tree induction. The proposed criterion applies an improved dynamic programming algorithm used in scheduling problems to establish an optimal number of fuzzy items for each condition attribute. Then, based on these fuzzy partitions, a fuzzy decision tree is constructed in a top-down recursive way. A comparative analysis using several traditional decision trees verify the feasibility of the proposed dynamic programming based fuzzy partition criterion. Furthermore, under the same framework of fuzzy decision trees, the proposed fuzzy partition solution can obtain a higher classification accuracy than some cases with the same amount of fuzzy items.

Rubrication of text information based on the voting of intellectual classifiers

The practical implementation of the concept of electronic government is one of the priorities of Russian state policy. The organization of effective interaction between authorities and citizens is an important element of this concept. In addition to providing public services, it should include the processing of electronic appeals (applications, complaints, suggestions, etc.). Research has shown that the speed and efficiency of appeal processing largely depend on the quality of determining the thematic rubric, i.e. solving the rubrication task. The analysis of citizens' appeals received by the e-mail and official websites of public authorities has revealed several specific features (small size, errors in the text, free presentation style, description of several problems) that do not allow the successful application of traditional approaches to their rubrication. To solve this problem, it has been proposed to use various methods of intellectual analysis of unstructured text data (in particular, fuzzy logical algorithms, fuzzy decision trees, fuzzy pyramidal networks, neuro-fuzzy classifi convolutional and recurrent neural networks). The article describes the conditions of the applicability of six intellectual classifiers proposed for rubricating the electronic citizens’ appeals. They are based on such factors as the size of the document, the degree of intersection of thematic rubrics, the dynamics of their thesauruses, and the amount of accumulated statistical information. For a situation where a specific model cannot make an unambiguous choice of a thematic rubric, it is proposed to use the classifier voting method, which can significantly reduce the probability of rubrication errors based on the weighted aggregation of solutions obtained by several models selected using fuzzy inference.

Pruning trees in C-fuzzy random forest

Pruning decision trees is the way to decrease their size in order to reduce classification time and improve (or at least maintain) classification accuracy. In this paper, the idea of applying different pruning methods to C-fuzzy decision trees and Cluster–context fuzzy decision trees in C-fuzzy random forest is presented. C-fuzzy random forest is a classifier which we created and we are improving. This solution is based on fuzzy random forest and uses C-fuzzy decision trees or Cluster–context fuzzy decision trees—depending on the variant. Five pruning methods were adjusted to mentioned kind of trees and examined: Reduced Error Pruning (REP), Pessimistic Error Pruning (PEP), Minimum Error Pruning (MEP), Critical Value Pruning (CVP) and Cost-Complexity Pruning. C-fuzzy random forests with unpruned trees and trees constructed using each of these pruning methods were created. The evaluation of created forests was performed on eleven discrete decision class datasets (forest with C-fuzzy decision trees) and two continuous decision class datasets (forest with Cluster–context fuzzy decision trees). The experiments on eleven different discrete decision class datasets and two continuous decision class datasets were performed to evaluate five implemented pruning methods. Our experiments show that pruning trees in C-fuzzy random forest in general reduce computation time and improve classification accuracy. Generalizing, the best classification accuracy improvement was achieved using CVP for discrete decision class problems and REP for continuous decision class datasets, but for each dataset different pruning methods work well. The method which pruned trees the most was PEP and the fastest one was MEP. However, there is no pruning method which fits the best for all datasets—the pruning method should be chosen individually according to the given problem. There are also situations where it is better to remain trees unpruned.

MULTI-MODEL APPROACH TO THE INTELLIGENT ANALYSIS OF CITIZENS’ APPEALS

Today, implementation of the concept of electronic government is one of the priority tasks of state policy in Russia. One of the elements of this concept is organizing effective interaction between authorities and citizens (the Government-to-Citizen model), which, besides providing public services, should include processing of electronic applications (applications, complaints, suggestions, etc.). In turn, the speed and efficiency of processing the incoming requests depends to a large extent on the quality of the definition of the corresponding thematic heading, i.e. solving the problem of rubrication (classification). An analysis of citizens’ appeals to e-mail and official websites of various government bodies revealed a number of specific features (small size, errors in the text, free style of presentation, description of several problems) that do not allow the successful application of traditional approaches to rubrication. To solve this problem, it was proposed to use various methods of mining unstructured text data (in particular, fuzzy-logical algorithms, fuzzy decision trees, fuzzy pyramidal networks, neuro-fuzzy classifier, convolutional and recurrent neural networks). This article describes a new approach to the analysis of electronic communications from citizens, based on the complex application of several rubrication models, which is distinguished by taking into account the degree of intersection of thematic headings, the dynamism of their thesauri and the volume of accumulated statistical information. For a situation where a specific model cannot make an unambiguous choice of a thematic heading, it is proposed to use the method of voting of classifiers, which can significantly reduce the probability of classification errors based on weighted aggregation of solutions obtained by several models.