Research On The Practical Application Of Different Scale Data Mining Based Algorithms

At present,the world has embraced the era of data information flow,and the mining of effective information in large-scale data is getting more and more attention.Although the study of data mining through decades of development has made a lot of research results,the application effects of different data mining techniques are not the same.For different research fields,the effects of different data mining model algorithms have always been a substantial topic for researchers.For example,in the aspect of image recognition in recent years,deep neural networks have been proved that their recognition accuracy is better than other recognition models.However,in the field of identifying small scale feature data samples,it is often supported that vector machines or random forest algorithms can have higher precision.Therefore,studying "such"different fields does not have a universal optimal model algorithm for data mining and application in different research fields.Consequently,in the era based on data flow information,this paper develops the practical application in diverse research application fields by the mining model algorithm based on various types of data.Besides,this paper will promote its conclusion to practical application,so as to provide some guidance for improving data mining in different fields.And the concrete research contents and results are shown as follows respectively:(1)The monitoring of land rock desertification based on TM remote sensing data has been studied.By deeply analyzing the characteristics of TM remote sensing data,a SOM neural network is established to classify and identify the obtained indexes of each characteristic.This model algorithm synthesizes the characteristic index of each TM remote sensing data,and tests through the simulation of the actual TM remote sensing data.The results showed that:the SOM neural network model based on the characteristic index of TM remote sensing data can effectively obtain the overall comprehensive evaluation of the study area.That is to say,the model can plan and develop the core distribution results and monitor the distribution of rocky desertification,the vegetation distribution and the watershed distribution in the research area in real time.It has practical guiding significance in urban construction development and planning,which can provide theoretical basis for the analysis of remote sensing image data of rocky desertification in the study area.(2)For the financial warning monitoring of the listed companies,this paper has studied the model and the entropy method that are both based on the principal component,and put forward the enterprise financial warning monitoring model based on the fusion of SOM network and BP network.Besides,the simulation verification of each model has been carried out,and the results showed the following aspects:(a)the traditional principal component model and the fusion entropy model based on principal component can make a certain distinction between non-ST and ST enterprises.However,the recognition rate of this model can be improved greatly;(b)as for the financial warning model based on the fusion of SOM network and BP network proposed in this paper,it can extract and filtrate all objects of ST samples and non-ST samples.Depending on SOM network,the initial training samples were extracted,which not only obtained the matching ratio of the optimal sample,but also constructed a complete financial alert model system.Furthermore,it can also construct a complete finical alert model system and the recognition accuracy rate is higher than the direct use of Logistic model,BP neural network model,SVM model,the principal component model,and the principal component fusion entropy model.(3)For the mining and analysis of the traffic flow data,the key traffic flow parameters in different environments are collected.A Van-Aerd model was established to calibrate the traffic road parameters and analyze the practical effects of different environments on the road.At the same time,the calculation principle of FCM clustering model is analyzed in detail for the features of identifying the practical traffic state.In this paper,the adaptive identification of optimal cluster number in various environments is established based on the effective function under each situation,and the actual traffic flow data is simulated.The results showed that:The model algorithm proposed in this paper can effectively divide the traffic state of the expressway.The state of various types of traffic flows can be reflected through the central value of different clusters.And a clear state type is given for the uncertainty degree of the traffic state,which lays a foundation for the actual traffic control and management.(4)In this paper,the chaotic system based on Logistic mapping is studied for the security techniques of the image and information under the data scale.On the basis of one-dimensional Logistic mapping,the dynamic equation of two-dimensional chaotic Logistic mapping is introduced,and the image security is simulated and analyzed based on the two-dimensional chaotic Logistic system.At the same time,in order to overcome the distortion of the two-dimensional chaotic Logistic system in image decryption process and to meet the requirement of high accuracy transmission of image local information,this paper puts forward the image encryption algorithm based on chaos and the wavelet transform.Finally,it simulates through the actual experiment and the results show the following aspects:(a)when the two dimensional Logistic chaotic system is used to encrypt the two-dimensional images,the total key space magnitude can reach the level of 1018.When it extends to the encryption of the three-dimensional images,the total key space magnitude can reach the level of 1054,which has relatively good security but can cause certain image distortion;(b)the two-dimensional Logistic chaotic system with the wavelet transform algorithm is not only safe,but also overcomes the distortion during the process of image encryption and decryption.When the decomposition layer of wavelet is 2 layers,the magnitude of its key space will reach the level of 10126.This not only can ensure that the partial information of the image is preserved,but also the security can be greatly improved.