Research And Implementation Of Lung Cancer Forecast System Based On ANN And GA

The neural network can predict to the complicated problem accurately, but easy to affect by over train and trained slowly. The hereditary algorithm is that a kind of the overall situation optimizes the algorithm of searching for, simple and in common use, the characteristic such as being stupid and excellent and strong, however, one degree of function of fitting of it changes very big. The main request for this method fits one degree of function must disappear in the minimal error, this has left a lot of free space for concrete realization scheme. The ones that combined the hereditary algorithm and neural network together and used and can raise the model on higher level are intelligible. This project has proposed that the improved hereditary algorithm combines the method of excavating that BP optimizes the neural network. The work of this text includes mainly:1. Have carried on detailed research to the hereditary algorithm and artificial neural networkIncluding hereditary algorithm prototype, the way of the hereditary algorithm code, go against the error and propagate the algorithm (BP algorithm) Have discussed to the pluses and minuses of hereditary algorithm and artificial neural network.2. Combine the method of excavating that BP optimizes the neural network after putting forward the improved hereditary algorithmHave consulted the combination method of hereditary algorithm and neural network at present: First, optimize the neural initial right value of network; Second, adopt the neural network method to evolve (evolving neural networks, is abbreviated as ENN), totally substitute BP to study with the hereditary algorithm, in order to avoid the defect of the descent method of the gradient. But both there is a greater defect in these two kinds of combination methods: As to the former, hereditary algorithm itself have early-maturing question of disappearing too, so method this can guarantee succeeding network is it can fall into some little area extremely to train still; As to the latter, search for because of hereditary algorithm own part ability weak characteristic is it evolve neural network need one big initial right value area of range to make, it is unsatisfactory that the complexity arising from this rises and makes the neural network trained out suffused with the ability of melting, meanwhile, the population of the hereditary algorithm calculates that often makes its training expenses much larger than the time expenses of BP algorithm.Because of above-mentioned analysis, this text has proposed that the improved hereditary algorithm combines BP and optimizes the neural network method. Neural network algorithm have easy to fall into some advantage most and restrain the slow shortcoming to feed forward type, this text has introduced BP in non-linear least square method and optimized algorithms, this algorithm has stronger some search ability, can improve the precision of training of the network, shorten and train time. Consider hereditary algorithm in the course of searching for constantly to may include optimum direction that solve is it search for space to change, it is much bigger compared with simple neural network algorithm to search for to the optimum probability that solves of the overall situation, searching for the respect and not so good as the neural network algorithm in the part , this text tries to divide two stages to use the hereditary algorithm to improve the network to train quality, carry on through hereditary algorithm thick to is it get one approximate solution of the overall situation to adjust at first, regard this as initial value, adopt hereditary algorithm and BP to optimize the neural network algorithm and run to train alternatively again, BP optimize neural network algorithm reach to it switches over to be can through appointed precision or some to is it is it realize to come to count with vigorous strides most hereditary algorithm, hereditary algorithm reach BP optimize neural network algorithm is it can operate through first intact hereditary operator to switch over (choose the operator, report to the leads hip after accomplish it operator, make a variation operator), is it realize to come, so each other with training as one's own initial right value or initial colony result of the other side, train repeatedly and alternatively, until reach the precision appointed or the most greatly and replace the step to count.Through carrying on the experiment in the typical data are collected, find that it is on good error than other algorithms participating in being compared that the improved hereditary algorithm combines BP and optimizes the neural network method (GBP algorithm) on the judging by accident rate, slightly slow in HAAM algorithm in restraining the speed. The experiment indicates, though HAAM algorithm is quicker than G algorithm in restraining the speed, it is suffused with performance of melting to drop to some extent ; And GBP algorithm, through using the hereditary algorithm alone on the initial stage of training and training to optimize the network algorithm to use with BP alternatively in middle period, can not only optimize the initial right value of the network but also adjust the initial right value used in BP algorithm each time effectively , meanwhile, BP algorithm strengthened some search ability of the hereditary algorithm too; Make the network train extremely some snack difficult to fall into, the network got has the better one to suffused with the ability of melting.3. Have developed the data and excavated the package because of GBP algorithm, and use the output of the Medical to predictOn combining BP and optimizing the neural network method foundation in improved hereditary algorithm put forward in this text, excavate the package after developing the data, this package is by the pretreatment of the data, the data excavate and excavate the result to show that three parts make up. The pretreatment of the data is mainly partly to deal with the data list in the database, the data lacked while expressing are mended completely, revise the wrong data; The data in the his-and-hers watches are excavated selectively according to the needs of user, if the attribute of the all right his-and-hers watches is arranged and write down the conduct to simplify. The data excavate every parameter that will excavate using partly to show to users in the form of window body, for it to set up according to the need, such as the precision of excavating, train number of times, etc. the most largely. Excavate the result and show some, will excavate the result to show to users in the form of data list. The project collect in order to train the data collection, with Cell type,Status, Karnofsky score, Months from Diagnosis, Age in years and Survival in days, in order to test the data collection , precipitation , in order to input attribute in order to output attribute, use this tool to excavate, to find that it is not people's mode corresponding relation known before to imply among the gene above. After excavating and analyzing actually, we have got the more ideal experimental result. Though it is difficult for the mode knowledge received by this way to understand, we can carry on the prediction analysis of the output of the Medical.