Based On Artificial Neural Network Prediction Of Population Research

In recent years, people constantly study population development rule, hope from thecomplex population find a rule to predict the population development in future, therebyestablishing a reasonable policy. But population growth is vulnerable to the birth rate,mortality and other objective factors and subjective factors such as the impact of populationpolicies, so the traditional methods for population prediction accuracy is often less than theexpected value. Artificial neural network is a nonlinear science, have very strong faulttolerance, adaptive and nonlinear mapping ability, to overcome the traditional artificialintelligence methods in terms of information processing defects, making it in the nervoussystem designed surface, pattern recognition, intelligent control, combinatorial optimization,prediction and other fields have been applied successfully.This paper uses three categories: artificial neural network back-propagation network, theRBF neural network, time series prediction method, study population prediction, topopulation prediction feature, considering the population prediction indexes, therebyreasonable prediction of population growth in numbers, for our country’s sustainabledevelopment provides a convenient. In BP network, in order to avoid network into localminimum and improve the convergence speed of the network, using the momentum andadaptive learning rate algorithm combining the national total population, populationprediction, using BP network of three layers, the number of the input neurons of 8, transportlayer neuron number1 in RBF neural network used in parameter is the basis function centerand variance and weight; in the AR model, using curve fitting and parameter estimationmethods ( such as nonlinear least square method ) for network training. The effect of nationaltotal population of each index is also used in the prediction of time series, set up the BPnetwork, RBF network and AR model prediction.By choosing 1990-2008year population index prediction, through the results show that, the total population of predictive value and the actual value of the basic agreement, BPpredictive value and population gross error is: 0.0046,0.0011, 0.0009,0.0035,0.RBF predictivevalue and population gross erroris:0.0012,0.00023,0.0062，0.0141,0.0056, 0.0079,0.0002,0.0005. So that the neural network used in population prediction is feasible and effective, hasa good prospect.