Study Of The Modeling Optimization Based On Neural Network And GA

The Ensemble Learning is a hotspot in the intelligent learning field. It has beenpaid attention from its' very beginning and has been used successfully in many fieldsbecause of its' good performance and usability. The combination of Ensemblealgorithms and unstable learners has shown impressive effect both in accuracy andstability. At the same time, partheno-genetic algorithm is a hotspot due to its' maturity.It is adored by so many researchers due to its' ability to solve combination problemsespecially problems in large scale.Based on the algorithms discussed above, this article does some useful study insome aspects. Something was done as below:(1) Some experiments were done to testify the validity of the Boosting algorithmbased on the study of the Ensemble algorithm. The experiments were basedon the UCI database, and the purpose is to prove the good performances ofBoosting in some aspects as: Serviceability;Stability. The ability ofparameter-optimization. At last, we discuss the Serviceability and"Over-fitting Problem" from the point of Margin., and show it in intuitionisticgraphics.(2) Based on the study above, we put forward a new method to ulteriorlyoptimize parameters of Boosting algorithm. Here, the methods we used are:Genetic-algorithm and LSE algorithm. And the simulation results show thatboth methods can improve the performance more.(3) As to partheno-genetic algorithm, we bring forward an improved algorithmbased on partheno-genetic algorithm to solve the CSP problem. A newmethod was brought forward to simplify the coding process, at the same time,an operator was introduced to help to find the best global solution. And thealgorithm has been testified to be a excellent one through the test in more thanone industrial enterprises. All proofs show that it is a good method to solvethe CSP problem both in the Cutting-Accuracy and Calculating-Efficiency.