Research On Estimation Of The Cost Of Equipment Software Based On Particle Swarm Optimization Tuned Support Vector Machine

As the significant guarantee of national defend safety,weapon equipment has directly relationship with national stabilizing development and peaceful and safety of residents.With the computer technique and advanced artificial intelligence technology are used in military area,the contest of equipment between nations has been from hardware strength shift towards computer software strength,equipment software has an increasing important role in modern equipment.At the same time,the rate of equipment project between software and hardware has been in steadily increase,the cost of equipment software has been the key part of modern equipment cost.Efficient and accurate equipment software cost estimation is the foundation to fix a price of equipment software and the assurance to accomplish the equipment software project successfully,which is favorable to the improvement of benefit in equipment software manufactures and the promotion of efficiency in defend fundsAt first,on the foundation of analyzing relative theory of equipment software cost estimation,this study analyses the notion and features of equipment software and the conception of equipment software cost estimation.Meantime,this study put forward to several problems about our nation equipment software cost estimation recently for its situation.Then,in order to improve the accuracy and scientificity of equipment software cost estimation,this paper analyzes the 17 contributory factors of equipment software cost from four aspects which are software itself,project progress,platform resource and employees,and constructs an index system through Delphi Technique and REM model.Based on the small-sample and high-dimension of equipment software cost statistics,this study analyzes serviceability of SVR in equipment software cost estimation and constructs SVR model.Meantime,to improve the accuracy and efficiency of SVR model,this study use particle swarm optimization(PSO)to tune penalty parameter C in SVR and parameter σ in Radial Basis Function and choose a best parameters combinationFinally,selecting twenty-nine statistics of flight management software in an aircraft,this study uses these statistics which dimensions have been reduced by Grey Relational Analysis to simulation research and test the effectiveness of PSO-SVR model.Through comparing simulation research result,this study concludes that PSO-SVR has improved the accuracy and efficiency of equipment software estimation.Estimation model of the cost of equipment software based on particle swarm optimization tuned support vector machine constructed by this paper have excellent accuracy and higher efficient,which is able to give a model and method reference to our nation equipment software cost estimation.