To obtain access to full text of journal and articles you must register!
- Article name
- RULES EXTRACTION FROM THE TRAINED NEURAL NETWORK IN THE CLASSIFICATION TASKS
- Authors
- Gridin V. N., , info@ditc.ras.ru, Federal State Institution of Science Center for Information Technology in the Design of Sciences, Moscow, Russia
Solodovnikov V. I., , info@ditc.ras.ru, Center for Information Technology in the Design of RAS, Odintsovo, Moscow region, Russia
- Keywords
- neural network / decision trees / logical conclusion / rules extraction / date mining
- Year
- 2017 Issue 4 Pages 49 - 54
- Code EDN
- Code DOI
- Abstract
- The issues of the joint use of neural network technologies with methods of logical deduction and decision making support during the rules construction in classification problems are considered. The analysis of algorithms for searching logical patterns in data is carried out. The prospects of using the neural network approach are grounded, where the rules are contained in the internal structure of the network: in weight coefficients, activation functions and neuron connections, but their formation occurs during the learning process. A description of combined algorithms for rule extraction from the trained neural networks and presenting the result in the form of a hierarchical, sequential structure of "if-then" rules is given. The decision making trees transformation into facts of the semantic network is considered.
- Text
- To obtain access to full text of journal and articles you must register!
- Buy